Professional Documents
Culture Documents
Documentation
Release 7.5.5.1
Luca Congedo
1 Introduction 1
2 Plugin Installation 3
2.1 Installation in Windows 32 bit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
2.2 Installation in Windows 64 bit . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.3 Installation in Ubuntu Linux . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.4 Installation in Debian Linux . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.5 Installation in Mac OS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
7 Semi-Automatic OS 193
7.1 Installation in VirtualBox . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 194
i
8.5 Errors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204
8.6 Various . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206
ii
CHAPTER 1
Introduction
Developed by Luca Congedo, the Semi-Automatic Classification Plugin (SCP) is a free open source plugin for
QGIS that allows for the semi-automatic classification (also known as supervised classification) of remote sensing
images. It provides several tools for the download of free images, the preprocessing, the postprocessing, and the
raster calculation (please see What can I do with the SCP? (page 206)). All the main tools are designed for parallel
processing.
The overall objective of SCP is to provide a set of intertwined tools for raster processing in order to make an
automatic workflow and ease the land cover classification, which could be performed also by people whose main
field is not remote sensing.
This user manual provides information about the Plugin Installation (page 3) of SPC and the The Interface of
SCP (page 19), with detailed information about all the functions. In addition, the Brief Introduction to Remote
Sensing (page 123) illustrates the basic concepts and definitions which are required for using the SCP.
Basic Tutorials (page 155) are available for learning the main functions of SCP.
You are kindly invited to contribute to SCP (see How to contribute to SCP (page 207)) and join the Facebook
group . Also, please read the Frequently Asked Questions (page 197).
For more information and tutorials visit the official site
1
Semi-Automatic Classification Plugin Documentation, Release 7.5.5.1
2 Chapter 1. Introduction
CHAPTER 2
Plugin Installation
The Semi-Automatic Classification Plugin requires the installation of GDAL, OGR, NumPy, SciPy and Matplotlib
(already bundled with QGIS).
This chapter describes the installation of the Semi-Automatic Classification Plugin for the supported Operating
Systems.
3
Semi-Automatic Classification Plugin Documentation, Release 7.5.5.1
• Run QGIS;
• From the main menu, select Plugins > Manage and Install Plugins;
• From the menu All, select the Semi-Automatic Classification Plugin and click the button Install
plugin;
TIP: in case of issues or an offline installation is required see How to install the plugin manually?
(page 198) and How to install the plugin from the official SCP repository? (page 198).
• The SCP should be automatically activated; however, be sure that the Semi-Automatic Classification Plu-
gin is checked in the menu Installed (the restart of QGIS could be necessary to complete the SCP
installation);
Now, the Semi-Automatic Classification Plugin is installed and a dock and a toolbar should be added to QGIS.
Also, a SCP menu is available in the Menu Bar of QGIS. It is possible to move the toolbar and the dock according
to your needs, as in the following image.
• Run QGIS;
• From the main menu, select Plugins > Manage and Install Plugins;
• From the menu All, select the Semi-Automatic Classification Plugin and click the button Install
plugin;
TIP: in case of issues or an offline installation is required see How to install the plugin manually?
(page 198) and How to install the plugin from the official SCP repository? (page 198).
• The SCP should be automatically activated; however, be sure that the Semi-Automatic Classification Plu-
gin is checked in the menu Installed (the restart of QGIS could be necessary to complete the SCP
installation);
Now, the Semi-Automatic Classification Plugin is installed and a dock and a toolbar should be added to QGIS.
Also, a SCP menu is available in the Menu Bar of QGIS. It is possible to move the toolbar and the dock according
to your needs, as in the following image.
The configuration of available RAM is recommended in order to reduce the processing time. From the SCP menu
In the Settings (page 107), set the Available RAM (MB) to a value that should be half of the system RAM.
For instance, if your system has 2GB of RAM, set the value to 1024MB.
• Press Enter and wait until the software is downloaded and installed.
Now, QGIS is installed.
• Run QGIS;
• From the main menu, select Plugins > Manage and Install Plugins;
• From the menu All, select the Semi-Automatic Classification Plugin and click the button Install
plugin;
TIP: in case of issues or an offline installation is required see How to install the plugin manually?
(page 198) and How to install the plugin from the official SCP repository? (page 198).
• The SCP should be automatically activated; however, be sure that the Semi-Automatic Classification Plu-
gin is checked in the menu Installed (the restart of QGIS could be necessary to complete the SCP
installation);
Now, the Semi-Automatic Classification Plugin is installed and a dock and a toolbar should be added to QGIS.
Also, a SCP menu is available in the Menu Bar of QGIS. It is possible to move the toolbar and the dock according
to your needs, as in the following image.
The configuration of available RAM is recommended in order to reduce the processing time. From the SCP menu
In the Settings (page 107), set the Available RAM (MB) to a value that should be half of the system RAM.
For instance, if your system has 2GB of RAM, set the value to 1024MB.
• Type in a terminal:
• Press Enter and wait until the software is downloaded and installed.
Now, QGIS is installed.
• Run QGIS;
• From the main menu, select Plugins > Manage and Install Plugins;
• From the menu All, select the Semi-Automatic Classification Plugin and click the button Install
plugin;
TIP: in case of issues or an offline installation is required see How to install the plugin manually?
(page 198) and How to install the plugin from the official SCP repository? (page 198).
• The SCP should be automatically activated; however, be sure that the Semi-Automatic Classification Plu-
gin is checked in the menu Installed (the restart of QGIS could be necessary to complete the SCP
installation);
Now, the Semi-Automatic Classification Plugin is installed and a dock and a toolbar should be added to QGIS.
Also, a SCP menu is available in the Menu Bar of QGIS. It is possible to move the toolbar and the dock according
to your needs, as in the following image.
The configuration of available RAM is recommended in order to reduce the processing time. From the SCP menu
In the Settings (page 107), set the Available RAM (MB) to a value that should be half of the system RAM.
For instance, if your system has 2GB of RAM, set the value to 1024MB.
• Visit the website KyngChaos and download the latest QGIS disk image (including Python and GDAL).
QGIS 3.16 is recommended at the moment.
WARNING: it is recommended to install only one versions of QGIS; for a clean QGIS installa-
tion, uninstall any previous version of QGIS and remove all the related directories.
• Follow the Readme instructions and clean up any previous path settings. Python must be installed prior to
GDAL and QGIS.
Now, QGIS is installed.
• Run QGIS;
• From the main menu, select Plugins > Manage and Install Plugins;
• From the menu All, select the Semi-Automatic Classification Plugin and click the button Install
plugin;
TIP: in case of issues or an offline installation is required see How to install the plugin manually?
(page 198) and How to install the plugin from the official SCP repository? (page 198).
• The SCP should be automatically activated; however, be sure that the Semi-Automatic Classification Plu-
gin is checked in the menu Installed (the restart of QGIS could be necessary to complete the SCP
installation);
Now, the Semi-Automatic Classification Plugin is installed and a dock and a toolbar should be added to QGIS.
Also, a SCP menu is available in the Menu Bar of QGIS. It is possible to move the toolbar and the dock according
to your needs, as in the following image.
The configuration of available RAM is recommended in order to reduce the processing time. From the SCP menu
In the Settings (page 107), set the Available RAM (MB) to a value that should be half of the system RAM.
For instance, if your system has 2GB of RAM, set the value to 1024MB.
In case of missing dependencies, you could get error messages such as the following example with matplotlib:
The SCP interface is composed of several parts that are described in detail in the following paragraphs.
The SCP menu allows for the selection of the main functions of the Main Interface Window (page 31), the Spectral
Signature Plot (page 112), and the Scatter Plot (page 117).
19
Semi-Automatic Classification Plugin Documentation, Release 7.5.5.1
• Online help: open the Online help in a web browser; also, a Facebook group is available for sharing
information and asking for help about SCP;
• About: information about the plugin;
• Show plugin: show all the SCP toolbars and dock if previously hidden;
The SCP dock allows for the the creation of ROIs (Regions Of Interest) and spectral signatures required for the
classification of a Band set (page 31). The Training input (page 23), created with SCP, stores the ROI polygons
and spectral signatures used for the land cover classification (see Tutorial 1: Your First Land Cover Classification
(page 155)).
ROIs are polygons used for the definition of the spectral characteristics of land cover classes. Spectral signatures
of classes are calculated from the ROIs or can be imported from other sources (see Import signatures (page 37)).
It is worth pointing out that classification is always based on spectral signatures.
SCP allows for the creation of temporary ROI polygons using a region growing algorithm or drawn manually with
the tools provided in the Working toolbar (page 27). Temporary ROI polygons can be saved in the Training input
(page 23) along with the spectral signatures of the ROI.
The Training input (page 23) is composed of a vector part (i.e. a shapefile) and a spectral signature part which are
independent. The attribute table of the vector contains four fields as in the following table.
Training input fields
TIP: The Training input (page 23) is managed by SCP, a temporary layer is added to QGIS but the
real file is saved and modified during the editing in SCP.
In SCP, land cover classes (and ROIs) are defined with a system of Classes (Class ID) and Macroclasses (Macro-
class ID) (see Classes and Macroclasses (page 134)) that are used for the classification process; each Macroclass
ID is related to a Macroclass Information (e.g. macroclass name) and each Class ID is related to a Class Informa-
tion (e.g. class name), but only Macroclass ID and Class ID are used for the classification process.
The use of the Macroclass ID or Class ID for classifications is defined with the option Use MC ID or C ID in
the Algorithm (page 71). It is worth highlighting that when using Macroclass ID all the spectral signatures are
evaluated separately and each pixel is classified with the corresponding MC ID (i.e. there is no combination of
signatures before the classification).
The classification can be performed for the entire image (see Algorithm (page 71)) or a part of it, creating a
Classification preview (page 30).
The SCP dock contains the following tabs:
The left side of SCP dock contains buttons for accessing the main functions of SCP.
3.2.1 Home
The tab Home contains the buttons for accessing the main functions of SCP:
Fig. 2: Home
SCP news
This section displays news about the SCP and related services. News are downloaded on startup (internet connec-
tion required). It can be enabled or disabled in the settings Dock (page 110).
It also contains the following buttons:
• User manual: open the link to the user manual;
• Ask a question: open the link to the online help;
• Support the SCP: open the link to support the SCP ;
This tool allows for the creation of the training input file (.scp ) required for storing ROIs and spectral signa-
tures. The training input file is created according to the characteristics of the active band set defined in Band set
(page 31). A new training input file should be created for every band set, unless the band sets have the same
spectral characteristics and coordinate reference system.
• : open a training input file; ROIs and spectral signatures are loaded in ROI & Signature list (page 24);
• : create an empty training input file (.scp); the vector part of the training input is loaded in QGIS;
also a backup file is created (a file .scp.backup in the same directory as the file .scp) when the QGIS
project is saved; to use the backup file simply rename it deleting the ending .backup extension and open
it as training input;
The tab ROI & Signature list displays the ROI polygons and spectral signatures contained in the training input
file. The tab ROI & Signature list is structured as tree list, where every ROI is grouped in the corresponding
Macroclass.
Changes in the ROI & Signature list are applied to the file Training input (page 23) only when the QGIS project
is saved (but there is also the option Autosave). ROIs can be edited, deleted and merged from this list.
If an item is a ROI polygon, double click the item to zoom to that ROI in the map. Items in the list can be
highlighted with the mouse left click.
TIP: ROIs and spectral signatures can be imported from other sources (see Import signatures
(page 37)) and exported (see Export signatures (page 36)).
– Name: Macroclass and Class Name; it can be edited with a single click;
– Type: type of the item:
• : merge highlighted spectral signatures or ROIs obtaining a new signature calculated as the average of
signature values for each band (covariance matrix is excluded);
• : calculate spectral signatures of highlighted ROIs using the active band set in Band set (page 31);
• : show the ROI spectral signature in the Spectral Signature Plot (page 112); spectral signature is
calculated from the Band set (page 31);
• : open the tab Export signatures (page 36) and export highlighted items;
ROI & Signature list is complementary to the Working toolbar (page 27) and it allows for saving ROIs to the
Training input (page 23) defining classes and macroclasses. A Band set (page 31) must be defined before the ROI
creation, and ROI polygons must be inside the area of the Band set.
• : delete the last saved ROI from the Training input (page 23);
• Autosave : if checked, automatically save the ROI & Signature list to the Training input (page 23)
every time a ROI is saved;
• Signature : if checked, while saving a ROI, the spectral signature thereof is calculated (from Band
set (page 31) pixels under ROI polygon) and saved to Training input (page 23) (calculation time depends on
the band number of the active band set in Band set (page 31));
• : save the temporary ROI to the Training input (page 23) using the defined classes and macroclasses;
ROI is displayed in the ROI & Signature list (page 24);
A right click on ROI & Signature list (page 24) allows for opening a menu containing several functions to manage
ROIs and spectral signatures.
• Zoom to: zoom to highlighted items (if ROI polygons) in the map;
• Change MC ID: edit the macroclass of highlighted items (using the value MC ID displayed in
ROI & Signature list (page 24)); if a macroclass is selected, the function is performed to all the included
items;
• Change color: select a color for the highlighted items; if a macroclass is selected, the function is
performed to all the included items;
• Merge items: merge highlighted spectral signatures or ROIs obtaining a new signature calculated as
the average of signature values for each band (covariance matrix is excluded); if a macroclass is selected,
the function is performed to all the included items;
• Calculate signatures: calculate spectral signatures of highlighted ROIs using the active band set in
Band set (page 31); if a macroclass is selected, the function is performed to all the included items;
• Delete items: delete highlighted ROIs and signatures; if a macroclass is selected, the function is
performed to all the included items;
• Add to spectral plot: show the ROI spectral signature in the Spectral Signature Plot (page 112);
spectral signature is calculated from the Band set (page 31); if a macroclass is selected, the function is
performed to all the included items;
• Add to scatter plot: add highlighted ROIs to the Scatter Plot (page 117); if a macroclass is selected,
the function is performed to all the included items;
• Export: open the tab Export signatures (page 36) and export highlighted items;
ROI options are useful for displaying pixel values or improving the creation of ROIs.
• Display : if the ROI creation pointer is active (see Working toolbar (page 27)), the pixel value of selected
– NDVI (Normalized Difference Vegetation Index); NDVI requires the near-infrared and red bands;
– EVI (Enhanced Vegetation Index); EVI requires the blue, near-infrared and red bands converted
to reflectance; wavelengths must be defined in the Band set (page 31);
– Custom; use the custom expression defined in the following line;
– : set a custom expression; expression is based on the Band set; bands are defined as
bandset#b + band number (e.g. bandset#b1 for the first band of the Band set); for example
NDVI for a Landsat image would be ( bandset#b4 - bandset#b3 ) / ( bandset#b4 +
bandset#b3 );
• Rapid ROI b. : if checked, temporary ROI is created with region growing using only one
Band set (page 31) band (i.e.region growing is rapider); the band is defined by the Band set number; if
unchecked, ROI is the result of the intersection between ROIs calculated on every band (i.e. region growing
is slower, but ROI is spectrally homogeneous in every band);
• Auto-plot: calculate automatically the temporary ROI spectral signature and display it in the Spectral
Signature Plot (page 112) (MC Name of this spectral signature is set tempo_ROI);
• Auto-refresh ROI: calculate automatically a new temporary ROI while Region growing parameters in
the Working toolbar (page 27) are being changed;
The Working toolbar allows for creating temporary ROIs and classification previews.
• : zoom the map to the extent of the active band set in Band set (page 31);
• RGB= : use the button to show/hide the active band set in Band set (page 31) in the map;
from the list select a Color Composite (page 133) that is applied to the Band set (page 31); new color
composites can be entered typing the band numbers separated by - or ; or , (e.g. RGB = 4-3-2 or RGB =
4;3;2 or RGB = 4,3,2);
• : display the input image stretching the minimum and maximum values according to cumulative count
of current map extent;
• : display the input image stretching the minimum and maximum values according to standard deviation
of current map extent;
http://www.youtube.com/watch?v=9jA_Rh8YfKQ
A temporary ROI is a temporary polygon displayed in the map, which can be saved permanently in the Training
input (page 23). A temporary ROI can be drawn manually or using a Region Growing Algorithm (page 134) (i.e.
the image is segmented around a pixel seed including spectrally homogeneous pixels).
• ROI: use the button to show/hide the temporary ROI and the Training input in the map;
• : activate the pointer to create a temporary ROI by drawing a polygon in the map; left click on the map
to define the ROI vertices and right click to define the last vertex closing the polygon; press the keyboard
button CTRL to add a multipart polygon; press the keyboard buttons CTRL + Z for removing the last
multipart polygon;
• : activate the pointer to create a temporary ROI using the region growing algorithm; left click on the
map for creating the ROI; right click on the map for displaying the spectral signature of a pixel of the active
band set in the Spectral Signature Plot (page 112); press the keyboard button CTRL and left click to add
a multipart polygon (new parts are not created if overlapping to other parts); press the keyboard buttons
CTRL + Z for removing the last multipart polygon; press the keyboard button CTRL and right click to plot
spectral signatures of the same pixel for all the band sets;
• : create a temporary ROI using the region growing algorithm at the same seed pixel as the previous
one; it is useful after changing the region growing parameters;
Region growing parameters: the following parameters are required for the ROI creation using a region growing algorithm on
• Dist : set the interval which defines the maximum spectral distance between the seed pixel
and the surrounding pixels (in radiometry unit);
• Min : set the minimum area of a ROI (in pixel unit); this setting overrides the Range
radius until the minimum ROI size is reached; if Rapid ROI on band is checked, then ROI
will have at least the size defined Min ROI size; if Rapid ROI on band is unchecked, then
ROI could have a size smaller than Min ROI size;
• Max : set the maximum width of a ROI (i.e. the side length of a square, centred at the seed
pixel, which inscribes the ROI) in pixel unit;
http://www.youtube.com/watch?v=CQqNtlOU6mI
Classification preview allows for displaying temporary classifications (i.e. classification previews). Classification
previews are useful for testing the algorithm in a small area of the Band set (page 31), before classifying the entire
image which can be time consuming (see Classification output (page 72)).
Classification preview is performed according to the parameters defined in Algorithm (page 71).
WARNING: ROIs and previews are performed on the active Band set (page 31).
In addition to the classification raster, an Algorithm raster (page 140) can be displayed, which is useful for assess-
ing the distance of a pixel classified as class X from the corresponding spectral signature X. In Classification
previews, black pixels are distant from the corresponding spectral signature (i.e. probably a new ROI, or spectral
signature, should be collected in that area) and white pixels are closer to the corresponding spectral signature (i.e.
probably the spectral signature identifies correctly those pixels).
After the creation of a new preview, old previews are placed in QGIS Layers inside a layer group named
Class_temp_group (custom name can be defined in Temporary group name (page 110)) and are deleted
when the QGIS session is closed.
WARNING: Classification previews are automatically deleted from disk when the QGIS session is
closed; a QGIS message (that can be ignored) could ask for the path of missing layers when opening
a previously saved project.
• : zoom the map to the extent of the last Classification preview (page 30);
• Preview: use the button to show/hide the last Classification preview (page 30) in the map;
• : activate the pointer for the creation of a Classification preview (page 30); left click the map to start
the classification process and display the classification preview; right click to start the classification process
and show the Algorithm raster (page 140) of the preview;
• : create a new Classification preview (page 30) centred at the same pixel as the previous one;
• T : change dynamically the classification preview transparency, which is useful for comparing the
classification to other layers;
• S : size of the preview in pixel unit (i.e. the side length of a square, centred at the clicked pixel);
• : remove from QGIS the classification previews that are archived in the Class_temp_group;
The Main Interface Window is composed of several tabs described in detail in the following paragraphs. Tabs can
be selected through the tree menu at the left side or from the SCP menu (page 19).
Image input in SCP is named band set. This tab allows for the definition of one or more band sets used as input
for classification and other tools.
Band sets are identified by numbers. The active band set (i.e. the tab selected in Band set definition (page 32)
with bold green name) is used as input for the tools in SCP dock (page 20) and Working toolbar (page 27). Other
SCP tools allow for the selection of band set numbers.
The Band set definition is saved with the QGIS project.
The tool Band set list (page 35) allows for the management of band sets.
This section allows for the selection of a multiband raster. If selected, raster bands are listed in the active band set.
• : select the input image from a list of multispectral images loaded in QGIS;
• : open one or more raster files that are added to the active band set and loaded in QGIS;
List of single band rasters already loaded in QGIS. It is possible to select one or more bands to be added to the
active band set.
• Filter : filter the list;
Definition of bands composing the band sets . The active band set is the tab selected with bold green name. It is
possible to add new band sets clicking the following button:
Click the in the tab to remove the corresponding band set. Band sets can be reordered dragging the tabs.
The Center wavelength of bands should be defined in order to use several functions of SCP. If the Center wave-
length of bands is not defined, the band number is used and some SCP tools will be disabled.
It is possible to define a multiplicative rescaling factor and additive rescaling factor for each band (for instance us-
ing the values in Landsat metadata), which are used on the fly (i.e. pixel value = original pixel
value * multiplicative rescaling factor + additive rescaling factor) during the
processing.
Every band set is defined with the following table:
• : sort automatically bands by name, giving priority to the ending numbers of name;
• Wavelength quick settings : rapid definition of band center wavelength for the following satellite sensors:
– ASTER;
– GeoEye-1;
– GOES;
– Landsat 8 OLI;
– Landsat 7 ETM+;
– Landsat 4-5 TM;
– Landsat 1, 2, and 3 MSS;
– MODIS;
– Pleiades;
– QuickBird;
– RapidEye;
– Sentinel-2;
– Sentinel-3;
– SPOT 4, 5, and 6;
– WorldView-2 and WorldView-3;
The tab Basic tools includes several tools for manipulating input data.
This tab allows for the definition of band weights that are useful for improving the spectral separability of materials
at certain wavelengths (bands). During the classification process, band values and spectral signature values are
multiplied by the corresponding band weights, thus modifying the spectral distances. A tab is displayed for every
Band set (page 31).
Band weight
Automatic weight
• Set weight : set the defined value as weight for all the highlighted bands in the table;
This tab allows for the management of band sets defined in Band set (page 31).
It is possible to order or remove multiple band sets at once. Double click the list to set the active band set (i.e. the
tab selected in Band set definition (page 32) with bold green name).
It is also possible to export and import multiple band sets saved as file .scpb .
• : create a RGB Color Composite (page 133) for highlighted band sets;
• : import a previously saved band set list from file (i.e. .scpb);
http://www.youtube.com/watch?v=1PclAtPjyZE
Export signatures
This tool allows for exporting the signatures highlighted in the ROI & Signature list (page 24).
• Export as SCP file : create a new .scp file and export highlighted ROIs and spectral signatures as SCP
file (* .scp);
• Export as shapefile : export highlighted ROIs (spectral signature data excluded) as a new shapefile (*
.shp);
• Export as CSV file : open a directory, and export highlighted spectral signatures as individual CSV files
(* .csv) separated by semicolon ( ; );
Import signatures
The tab Import signatures allows for importing spectral signatures from various sources.
This tool allows for importing spectral signatures from various sources: a previously saved Training input
(page 23) (.scp file); a USGS Spectral Library (.asc file); a previously exported CSV file. In case of USGS
Spectral Library, the library is automatically sampled according to the image band wavelengths defined in the
Band set (page 31), and added to the ROI & Signature list (page 24);
• Select a file : open a file to be imported in the Training input (page 23);
Import vector
This tool allows for importing a vector (shapefile or geopackage), selecting the corresponding fields of the Training
input (page 23).
• Calculate sig.: if checked, the spectral signature is calculated while the ROI is saved to Training input
(page 23);
• Import vector : import all the vector polygons as ROIs in the Training input (page 23);
The tab Download USGS Spectral Library allows for the download of the USGS spectral library (Clark, R.N.,
Swayze, G.A., Wise, R., Livo, E., Hoefen, T., Kokaly, R., Sutley, S.J., 2007, USGS digital spectral library splib06a:
U.S. Geological Survey, Digital Data Series 231).
The libraries are grouped in chapters including Minerals, Mixtures, Coatings, Volatiles, Man-Made, Plants, Vege-
tation Communities, Mixtures with Vegetation, and Microorganisms. An internet connection is required.
• Select a chapter : select one of the library chapters; after the selection, chapter libraries are shown in
Select a library;
• Select a library : select one of the libraries; the library description is displayed in the frame Library
description;
• Import spectral library : download the library and add the sampled spectral signature to the ROI &
Signature list (page 24) using the parameters defined for class and macroclass; the library is automatically
sampled according to the image band wavelengths defined in the active band set in Band set (page 31), and
added to the ROI & Signature list (page 24);
TIP: Spectral libraries downloaded from the USGS Spectral Library can be used with
Minimum Distance or Spectral Angle Mapping algorithms, but not Maximum Likelihood be-
cause this algorithm needs the covariance matrix that is not included in the spectral libraries.
LCS threshold
This tab allows for setting the signature thresholds used by Land Cover Signature Classification (page 139). All
the signatures contained in the Training input (page 23) are listed; also, signature thresholds are saved in the
Training input (page 23).
Overlapping signatures (belonging to different classes or macroclasses) are highlighted in orange in the table LC
Signature threshold; the overlapping check is performed considering MC ID or C ID according to the setting Use
MC ID C ID in Algorithm (page 71). Overlapping signatures sharing the same ID are not highlighted.
LC Signature threshold
• : show the ROI spectral signature in the Spectral Signature Plot (page 112); spectral signature is
calculated from the Band set (page 31);
Automatic thresholds
Set thresholds automatically for highlighted signatures in the table LC Signature threshold; if no signature is
highlighted, then the threshold is applied to all the signatures.
• Min Max : set the threshold based on the minimum and maximum of each band;
• 𝜎* : set an automatic threshold calculated as (band value + (𝜎 * v)), where 𝜎 is the standard
deviation of each band and v is the defined value;
• From ROI : set the threshold using the temporary ROI pixel values, according to the following checkboxes:
• From pixel : set the threshold by clicking on a pixel, according to the following checkboxes:
This tab allows for the automatic creation of ROIs, useful for the rapid classification of multi-temporal images, or
for accuracy assessment. Given a list of point coordinates and ROI options, this tool performs the region growing
of ROIs. Created ROIs are automatically saved to the Training input (page 23). The active band set in Band set
(page 31) is used for calculations.
• Number of points : set a number of points that will be created when Create points is clicked;
• inside grid : if checked, the band set area is divided in cells where the size thereof is defined in
the combobox (image unit, usually meters); points defined in Number of random points are created
randomly within each cell;
• min distance : if checked, random points have a minimum distance defined in the combobox
(image unit, usually meters); setting a minimum distance can result in fewer points than the number defined
in Number of points;
• Create points : create random points inside the band set area;
• stratified for the values of the first band of the band set min distance : if checked, create
random points inside the values defined in the expression calculated for the first band of the defined band
set; the expression must include the variable raster ; multiple expressions can be entered separated by
semicolon ( ; ) but the total number of stratified points is the same as the defined Number of points;
• Point coordinates and ROI definition: table containing the following fields;
– X : point X coordinate (float);
– Y : point Y coordinate (float);
– MC ID: ROI Macroclass ID (int);
– MC Name: ROI Macroclass Name (text);
– C ID: ROI Class ID (int);
– C Name: ROI Class Name (text);
– Min : the minimum area of a ROI (in pixel unit);
– Max : the maximum width of a ROI (in pixel unit);
– Dist : the interval which defines the maximum spectral distance between the seed pixel and the
surrounding pixels (in radiometry unit);
– Rapid ROI band : if a band number is defined, ROI is created only using the selected band,
similarly to Rapid ROI band in ROI & Signature list (page 24) ;
• : add a new row to the table; all the table fields must be filled for the ROI creation;
• : import a point list from text file or a point shapefile to the table; in case of text file, every line must
contain values separated by tabs of X, Y, MC ID, MC Name, Class ID, C Name, Min, Max, Dist, and
optionally the Rapid ROI band; in case of shapefile, only point coordinates are imported;
• Calculate sig.: if checked, the spectral signature is calculated while the ROI is saved to Training input
(page 23);
• RUN : start the ROI creation process for all the points and save ROIs to the Training input (page 23);
RGB list
This tab allows for managing the RGB Color Composite (page 133) used in the list RGB= of the Image control
(page 29).
RGB list
Automatic RGB
• Band combinations : add the combinations of all bands (i.e. permutation) to the RGB list (page 42)
(e.g. 1-2-3, 1-2-4, . . . , 3-2-1);
Signature threshold
This tab allows for the definition of a classification threshold for each spectral signature. All the signatures
contained in the Training input (page 23) are listed. Thresholds defined in this tool are applied to classification
only if Threshold value in Algorithm (page 71) is 0.
This is useful for improving the classification results, especially when spectral signatures are similar. Thresholds
of signatures are saved in the Training input (page 23).
If threshold is 0 then no threshold is applied and all the image pixels are classified. Depending on the selected
Algorithm (page 71) the threshold value is evaluated differently:
• for Minimum Distance, pixels are unclassified if distance is greater than threshold value;
• for Maximum Likelihood, pixels are unclassified if probability is less than threshold value (max 100);
• for Spectral Angle Mapping, pixels are unclassified if spectral angle distance is greater than threshold value
(max 90).
Signature threshold
Automatic thresholds
• Set threshold : set the defined value as threshold for all the highlighted signatures in the table;
• Set threshold = 𝜎 * : for all the highlighted signatures, set an automatic threshold calculated as
the distance (or angle) between mean signature and (mean signature + (𝜎 * v)), where 𝜎 is the standard
deviation and v is the defined value; currently works for Minimum Distance and Spectral Angle Mapping;
The tab Download products includes the tools for searching and downloading free remote sensing images.
Also, automatic conversion to reflectance of downloaded bands is available.
An internet connection is required and free registration could be required depending on the download service.
This tool allows for searching and downloading:
• Landsat Satellites (page 127) images (from 1 MSS to 8 OLI) acquired from the 80s to present days;
• Sentinel-1 Satellites (page 132) images (GRD) acquired from 2014 to present days;
• Sentinel-2 Satellite (page 128) images (Level-1C and Level-2A) acquired from 2015 to present days;
• Sentinel-3 Satellite (page 129) images (OLCI Level-1B OL_1_EFR) acquired from 2016 to present days;
• ASTER Satellite (page 129) images (Level 1T) acquired from 2000 to present days;
• MODIS Products (page 130) images (MOD09GQ, MYD09GQ, MOD09GA, MYD09GA, MOD09Q1,
MYD09Q1, MOD09A1, MYD09A1) acquired from 2000 to present days;
• GOES Products (page 131) images (GOES 16 and GOES 17 Full Disk L1b Radiance) acquired from 2017
to present days;
For Landsat, ASTER, and MODIS the search is performed through the CMR Search API developed by NASA.
Landsat images are freely available through the services: EarthExplorer , Google Cloud Storage Landsat . The
ASTER L1T and MODIS products are retrieved from the online Data Pool, courtesy of the NASA Land Processes
Distributed Active Archive Center (LP DAAC), USGS/Earth Resources Observation and Science (EROS) Center,
Sioux Falls, South Dakota, https://lpdaac.usgs.gov/data_access/data_pool.
GOES products are downloaded through Amazon AWS Open Data .
For Sentinel-1, Sentinel-2, and Sentinel-3 the search is performed through the Copernicus Open Access Hub
(using the Data Hub API ). Sentinel-2 are downloaded from Google Cloud Storage if available.
Please notice that Sentinel products older than 12 months may be archived in Long Term Archives ) and should
be requested through the Copernicus Open Access Hub .
This tool attempts to download images first from Amazon Web Services and Google Earth Engine ; only if images
are not available, the download is performed through the services that require to login.
Login data
Login https://ers.cr.usgs.gov
For Landsat images USGS EROS credentials (https://ers.cr.usgs.gov) are required for downloads from EarthEx-
plorer . Login using your USGS EROS credentials or register for free at https://ers.cr.usgs.gov/register .
Login https://urs.earthdata.nasa.gov
For ASTER and MODIS images EOSDIS Earthdata credentials (https://urs.earthdata.nasa.gov ) are required for
download. Login using your EOSDIS Earthdata credentials or register for free at https://urs.earthdata.nasa.gov/
users/new . Before downloading ASTER and MODIS images, you must approve LP DAAC Data Pool clicking
the following link https://urs.earthdata.nasa.gov/approve_app?client_id=ijpRZvb9qeKCK5ctsn75Tg
Login Sentinels
Search
Search parameters
Define the search area by entering the coordinates (longitude and latitude) of an Upper Left (UL) point and
Lower Right (LR) point and select a product to search. Optional settings are date of acquisition, maximum
cloud cover, number of results (the less the results, the faster is the query).
The definition of a search area is required before searching the images. For Sentinel-2 it is possible to avoid the
definition of a search by entering a granule name in the Advanced search .
• : define a search area by left click to set the UL point and right click to set the LR point; the search
area is displayed in the map;
• Products : set the search product;
• Date from : set the starting date of acquisition;
• to : set the ending date of acquisition;
• Advanced search : some services (e.g. Sentinel-2) allows for entering search parameters; for instance,
it is possible to enter the name of a granule (e.g. 33TTG) without defining a search area in order to get the
results of that granule; Sentinel-2 level 2A can be filtered entering S2A_MSIL2A* ;
• Find : find the products in the search area; results are displayed inside the table in Product list
(page 48); results are added to previous results;
• Add OpenStreetMap to the map: this button allows for the display of OpenStreetMap tiles (© Open-
StreetMap contributors) in the QGIS map as described in https://wiki.openstreetmap.org/wiki/QGIS . The
cartography is licensed as CC BY-SA (Tile Usage Policy );
TIP: Search results (and the number thereof) depend on the defined area extent and the range of
dates. In order to get more results, perform multiple searches defining smaller area extent and
narrow acquisition dates (from and to).
Product list
The table Product list contains the results of the search. Click on any item (highlight) to display the image preview
thereof (not all the products include image previews). Results are saved with the QGIS project.
• Product list : found products are displayed in this table, which includes the following fields;
• : display preview of highlighted images in the map; preview is roughly georeferenced on the fly (not
all the products include a preview);
Download
Download the products in the Product list (page 48). During the download it is recommended not to interact with
QGIS.
• Only if preview in Layers: if checked, download only those images listed in Product list (page 48)
which are also listed in the QGIS layer panel;
• Preprocess images: if checked, bands are automatically converted after the download, according to the
settings defined in Landsat (page 53);
• Load bands in QGIS: if checked, bands are loaded in QGIS after the download;
• RUN : start the download process of all the products listed in Product list (page 48);
Download options
This tab allows for the selection of single bands for Landsat, Sentinel-2, and Sentinel-3 images. Depending on the
download service, single band download could be unavailable.
• Ancillary data: if checked, the metadata files are selected for download;
3.4.4 Preprocessing
The tab Preprocessing provides several tools for data manipulation which are useful before the actual clas-
sification process.
ASTER
This tab allows for the conversion of ASTER L1T images to the physical measure of Top Of Atmosphere re-
flectance (TOA), or the application of a simple atmospheric correction using the DOS1 method (Dark Object
Subtraction 1), which is an image-based technique (for more information about conversion to TOA and DOS1
correction, see Image conversion to reflectance (page 148)).
Once the input is selected, available bands are listed in the metadata table.
ASTER conversion
• Apply DOS1 atmospheric correction: if checked, the DOS1 Correction (page 149) is applied to all the
bands;
• Use value as NoData : if checked, pixels having NoData value are not counted during conversion
and the DOS1 calculation of DNmin; it is useful when image has a black border (usually pixel value = 0);
• Create Band set and use Band set tools: if checked, bands are added to the active Band set after the
conversion; also, the Band set is processed according to the tools checked in the Band set (page 31);
• Add bands in a new Band set: if checked, bands are added to a new empty Band set after the conversion;
Metadata
All the bands found in the Select file ASTER L1T are listed in the table Metadata. For information about Metadata
fields visit the ASTER page .
• Earth sun distance : Earth Sun distance in astronomical units (automatically calculated if Date is filled;
• RUN : select an output directory and start the conversion process; only bands listed in the table Meta-
data are converted; converted bands are saved in the output directory with the prefix RT_ and automatically
loaded in QGIS;
GOES
This tab allows for the conversion of GOES images provided in netCDF4 (.nc) format inside a directory.
GOES conversion
• Directory containing GOES bands : open a directory containing GOES bands; names of bands must
end with the corresponding number;
• Use value as NoData : if checked, pixels having NoData value are not counted during conver-
sion; it is useful when image has a black border (usually pixel value = 0);
• Create Band set and use Band set tools: if checked, bands are added to the active Band set after the
conversion; also, the Band set is processed according to the tools checked in the Band set (page 31);
• Add bands in a new Band set: if checked, bands are added to a new empty Band set after the conversion;
Metadata
All the bands found in the Directory containing GOES bands are listed in the table Metadata.
• RUN : select an output directory and start the conversion process; only bands listed in the table Meta-
data are converted; converted bands are saved in the output directory with the prefix RT_ and automatically
loaded in QGIS;
Landsat
This tab allows for the conversion of Landsat 1, 2, and 3 MSS and Landsat 4, 5, 7, and 8 images from DN
(i.e. Digital Numbers) to the physical measure of Top Of Atmosphere reflectance (TOA), or the application of
a simple atmospheric correction using the DOS1 method (Dark Object Subtraction 1), which is an image-based
technique (for more information about the Landsat conversion to TOA and DOS1 correction, see Image conversion
to reflectance (page 148)).
Once the input is selected, available bands are listed in the metadata table.
Pan-sharpening is also available; for more information read Pan-sharpening (page 145). The panchromatic band
is preprocessed only if pansharpening is checked.
WARNING: For the best spectral precision you should download the Landsat Level-2 Data Products
(Surface Reflectance) from https://earthexplorer.usgs.gov .
• Directory containing Landsat bands : open a directory containing Landsat bands; names of Land-
sat bands must end with the corresponding number; if the metadata file is included in this directory then
Metadata (page 55) are automatically filled;
• Select MTL file : if the metadata file is not included in the Directory containing Landsat bands,
select the path of the metadata file in order to fill the Metadata (page 55) automatically;
• Brightness temperature in Celsius: if checked, convert brightness temperature to Celsius (if a Landsat
thermal band is listed in Metadata (page 55)); if unchecked temperature is in Kelvin;
• Apply DOS1 atmospheric correction: if checked, the DOS1 Correction (page 149) is applied to all the
bands (thermal bands excluded);
• Use value as NoData : if checked, pixels having NoData value are not counted during conversion
and the DOS1 calculation of DNmin; it is useful when image has a black border (usually pixel value = 0);
• Perform pan-sharpening: if checked, a Brovey Transform is applied for the Pan-sharpening (page 145)
of Landsat bands;
• Create Band set and use Band set tools: if checked, bands are added to the active Band set after the
conversion; also, the Band set is processed according to the tools checked in the Band set (page 31);
• Add bands in a new Band set: if checked, bands are added to a new empty Band set after the conversion;
Metadata
All the bands found in the Directory containing Landsat bands are listed in the table Metadata. If the Landsat
metadata file (a .txt or .met file with the suffix MTL) is provided, then Metadata are automatically filled. For
information about Metadata fields read this page and this one .
• Earth sun distance : Earth Sun distance in astronomical units (automatically calculated if Date is filled;
• RUN : select an output directory and start the conversion process; only bands listed in the table Meta-
data are converted; converted bands are saved in the output directory with the prefix RT_ and automatically
loaded in QGIS;
MODIS
This tab allows for the conversion of MODIS images to .tif format, and the reprojection to WGS 84.
Once the input is selected, available bands are listed in the metadata table.
MODIS conversion
• Reproject to WGS 84: if checked, reproject bands to WGS 84, required for use in SCP;
• Use value as NoData : if checked, pixels having NoData value are not counted during conversion
and the DOS1 calculation of DNmin; it is useful when image has a black border (usually pixel value = -999);
• Create Band set and use Band set tools: if checked, bands are added to the active Band set after the
conversion; also, the Band set is processed according to the tools checked in the Band set (page 31);
• Add bands in a new Band set: if checked, bands are added to a new empty Band set after the conversion;
Metadata
All the bands found in the Select file MODIS are listed in the table Metadata. For information about Metadata
fields visit the MODIS page .
• ID : ID of the image;
• RUN : select an output directory and start the conversion process; only bands listed in the table Meta-
data are converted; converted bands are saved in the output directory with the prefix RT_ and automatically
loaded in QGIS;
Sentinel-1
This tab allows for the conversion of Sentine-1 images, in particular Sentinel-1 Satellites (page 132) images
(GRD) VV and VH polarizations. For the processing method please read Sentinel-1 preprocessing (page 132).
These processes are performed through the SNAP Graph Processing Tool (GPT) and a graph file .xml . Therefore,
SCP requires ESA SNAP to perform the preprocessing. For the installation of ESA SNAP, please visit this page
http://step.esa.int/main/download/snap-download/ .
The path to the SNAP gpt executable must be entered in settings Processing (page 110).
Sentinel-1 conversion
• Select SNAP xml graph (optional) : select a custom graph file (.xml) to be used for preprocessing;
predefined variables for the xml graph can be: $input for the input file, $polarization for VV
or VH polarization if only one polarization is checked, $output for the output file ($outputVV and
$outputVH if both polarizations are checked);
• Raster projection as Band set : if checked, project the raster to the same coordinate reference
system as a band set (default WGS 84);
• Use value as NoData : if checked, pixels having NoData value are not counted during conver-
sion;
• Create Band set and use Band set tools: if checked, bands are added to the active Band set after the
conversion; also, the Band set is processed according to the tools checked in the Band set (page 31);
• Add bands in a new Band set: if checked, bands are added to a new empty Band set after the conversion;
Metadata
All the bands found in the Directory containing Landsat bands are listed in the table Metadata. If the Landsat
metadata file (a .txt or .met file with the suffix MTL) is provided, then Metadata are automatically filled. For
information about Metadata fields read this page and this one .
• Earth sun distance : Earth Sun distance in astronomical units (automatically calculated if Date is filled;
• RUN : select an output directory and start the conversion process; only bands listed in the table Meta-
data are converted; converted bands are saved in the output directory with the prefix RT_ and automatically
loaded in QGIS;
Sentinel-2
This tab allows for the conversion of Sentinel-2 images Level-1C to the physical measure of Top Of Atmosphere
reflectance (TOA), or the application of a simple atmospheric correction using the DOS1 method (Dark Object
Subtraction 1), which is an image-based technique (for more information about conversion to TOA and DOS1
correction, see Image conversion to reflectance (page 148)). This tool can also convert Sentinel-2 images Level-
2A from DN to reflectance values.
Once the input is selected, available bands are listed in the metadata table. Bands with 20m spatial resolution are
resampled to 10m resolution without changing the original pixel value (i.e. one 20m pixel is divided in four 10m
pixels with the same value).
WARNING: For the best spectral precision you should download the Sentinel-2 Level-2A Products
(Surface Reflectance) or use the official SNAP tool for atmospheric correction (see http://step.esa.int).
Sentinel-2 conversion
• Directory containing Sentinel-2 bands : open a directory containing Sentinel-2 bands; names of
Sentinel-2 bands must end with the corresponding number; if the metadata file is included in this direc-
tory then Metadata (page 60) are automatically filled;
• Select metadata file : select the metadata file which is a .xml file whose name contains
MTD_MSIL1C); this is optional and can be generally ignored;
• Apply DOS1 atmospheric correction: if checked, the DOS1 Correction (page 149) is applied to all the
bands;
• Use value as NoData : if checked, pixels having NoData value are not counted during conversion
and the DOS1 calculation of DNmin; it is useful when image has a black border (usually pixel value = 0);
• Preprocess bands 1, 9, 10: if checked, preprocess also bands 1, 9, 10 (i.e. 60m bands);
• Create Band set and use Band set tools: if checked, bands are added to the active Band set after the
conversion; also, the Band set is processed according to the tools checked in the Band set (page 31);
• Add bands in a new Band set: if checked, bands are added to a new empty Band set after the conversion;
TIP: in order to convert L2A images you should move all the .jp2 files inside the same di-
rectory and rename the files with the band number in the ending of the name (e.g. from
name_02_10m.jp2 to name_02.jp2).
Metadata
All the bands found in the Directory containing Sentinel-2 bands are listed in the table Metadata. If the Sentinel-2
metadata file (a .xml file whose name contains MTD_MSIL1C) is provided, then Metadata are automatically filled.
For information about Metadata fields read this informative page .
• RUN : select an output directory and start the conversion process; only bands listed in the table Meta-
data are converted; converted bands are saved in the output directory with the prefix RT_ and automatically
loaded in QGIS;
Sentinel-3
This tab allows for the conversion of Sentinel-3 images (OL_1_EFR) to the physical measure of Top Of Atmo-
sphere reflectance (TOA), or the application of a simple atmospheric correction using the DOS1 method (Dark
Object Subtraction 1), which is an image-based technique (for more information about conversion to TOA and
DOS1 correction, see Image conversion to reflectance (page 148)). The following ancillary data are required for
conversion: instrument_data.nc , geo_coordinates.nc , tie_geometries.nc .
Once the input is selected, available bands are listed in the metadata table.
WARNING: Sentinel-3 bands are reprojected to WGS 84 coordinate system using a sample of pixels
from the file geo_coordinates.nc . For the best precision you should use the official SNAP tool (see
http://step.esa.int).
Sentinel-3 conversion
• Directory containing Sentinel-3 bands : open a directory containing Sentinel-3 bands; names of
Sentinel-2 bands must end with the corresponding number; ancillary data required for conversion must
be in the same directory;
• Apply DOS1 atmospheric correction: if checked, the DOS1 Correction (page 149) is applied to all the
bands;
• Use value as NoData : if checked, pixels having NoData value are not counted during conversion
and the DOS1 calculation of DNmin; it is useful when image has a black border (usually pixel value = 0);
• Create Band set and use Band set tools: if checked, bands are added to the active Band set after the
conversion; also, the Band set is processed according to the tools checked in the Band set (page 31);
• Add bands in a new Band set: if checked, bands are added to a new empty Band set after the conversion;
Metadata
All the bands found in the Directory containing Sentinel-3 bands are listed in the table Metadata.
• RUN : select an output directory and start the conversion process; only bands listed in the table Meta-
data are converted; converted bands are saved in the output directory with the prefix RT_ and automatically
loaded in QGIS;
This tab allows for cutting several image bands at once, using a rectangle defined with point coordinates or a
boundary defined with a vector.
• Select input band set : select the input Band set (page 31) to be clipped;
• Use value as NoData : if checked, set the value for NoData pixels (e.g. pixels outside the clipped
area);
• Output name prefix : set the prefix for output file names (default is clip);
Clip coordinates
Set the Upper Left (UL) and Lower Right (LR) point coordinates of the rectangle used for clipping; it is possible
to enter the coordinates manually. Alternatively use a vector.
• : define a clip area by drawing a rectangle in the map; left click to set the UL point and right click to
set the LR point; the area is displayed in the map;
• Use vector for clipping : if checked, use the selected vector (already loaded in QGIS) for clipping;
UL and LR coordinates are ignored;
• Use vector field for output name : if checked, a vector field is selected for clipping while iterating
through each vector polygon and the corresponding field value is added to the output name;
• Use temporary ROI for clipping: if checked, use a Temporary ROI (page 29) for clipping; UL and LR
coordinates are ignored;
• RUN : choose an output destination and clip rasters in the selected Band set;
Cloud masking
This tool allows for cloud masking, based on the values of a raster mask, creating an output masked band for each
band of the Band set (page 31). NoData is set in all the bands of the Band set (page 31) for pixels corresponding
to clouds.
• Select input band set : select the input Band set (page 31) to be masked;
• Select the classification : select a classification raster (already loaded in QGIS) which contains a cloud
class;
• Mask class values : set the class values to be masked; class values must be separated by , and - can
be used to define a range of values (e.g. 1, 3-5, 8 will select classes 1, 3, 4, 5, 8); if the text is red then
the expression contains errors;
• Use buffer of pixel size : if checked, a buffer is created for masked area, corresponding to the
defined number of pixels; this can be useful to dilate masked area in case the mask doesn’t cover the thinner
border of clouds;
• Output NoData value : set the value of NoData pixels corresponding to clouds;
• Output name prefix : set the prefix for output file names (default is mask);
This tool allows for the mosaic of band sets, merging the corresponding bands of two or more band sets defined
in the Band set (page 31). An output band is created for every corresponding set of bands in the band sets.
NoData values of one band set are replaced by the values of the other band sets.
• Band set list : list if band sets defined in the Band set (page 31); in case of overlapping images, the
pixel values of the first band set in the list are assigned.
• Use value as NoData : if checked, set the value of NoData pixels, ignored during the calculation;
• Output name prefix : set the prefix for output file names (default is mosaic);
This tab allows for reprojecting bands in a Band set (page 31) and perform other raster operations.
• Select input band set : select the input Band set (page 31) to be reprojected;
• Align to raster same extent as reference : if Align to raster is checked, use the same
coordinate reference system and pixel size as the selected reference; if same extent as reference is checked,
the output extent will be the same as the reference;
• Use EPSG code X resolution Y resolution : if checked, use the EPSG code for
defining the output coordinate reference system with X resolution and Y resolution;
• Resample pixel factor : if checked, new pixel size will be the original size times this factor;
• Resampling method : select the resampling method; options are nearest_neighbour, average,
sum, maximum, minimum, mode, median, first_quartile, third_quartile;
• Output type : select the output raster type; options are Auto (same as input), Float32, Int32,
UInt32, Int16, UInt16, Byte;
• Change output NoData value : if checked, set a new value for NoData pixels;
• Output name prefix : set the prefix for output file names (default is reproj);
Raster list
• Output name prefix : set the prefix for output file names (default is split);
• RUN : choose the output destination and stack selected rasters; output is automatically loaded in
QGIS;
Stack all the bands in a Band set (page 31) into a multiband raster.
• Select input band set : select the input Band set (page 31) to be stacked;
Vector to raster
• Use the value field of the vector : if checked, the selected field is used as attribute for the conver-
sion; pixels of the output raster have the same values as the vector attribute;
• Use constant value : if checked, the polygons are converted to raster using the selected constant
value;
• Select the type of conversion : select the type of conversion between Center of pixels and All pixels touched:
– Center of pixels: during the conversion, vector is compared to the reference raster; output raster
pixels are attributed to a polygon if pixel center is within that polygon;
– All pixels touched: during the conversion, vector is compared to the reference raster; output raster
pixels are attributed to a polygon if pixel touches that polygon;
• Select the reference raster : select a reference raster; pixels of the output raster have the same size and
alignment as the reference raster;
• Same extent as reference raster: the output raster will have the same extent as reference raster;
• RUN : choose the output destination and start the conversion to raster;
The tab Band processing provides several functions that can be applied to the Band set (page 31).
Band combination
This tab allows for the combination of bands loaded in a Band set (page 31). This tool is intended for com-
bining classifications in order to get a raster where each value corresponds to a combination of class values. A
combination raster is produced as output and the area of each combination is reported in an text file.
• Select input band set (of classifications) : select the input Band set (page 31);
• RUN : choose the output destination and start the calculation; also, the details about the combinations
are displayed in the tab Output and saved in a .txt file in the output directory;
Classification
This tab allows for the classification of the Band set (page 31) using the spectral signatures checked in ROI &
Signature list (page 24). Several classification options are set in this tab which affect the classification process
also during the Classification preview (page 30).
This tool allows for the use of the following algorithms: Minimum Distance, Maximum Likelihood, Spectral
Angle Mapping. Random Forest is available in the Random Forest (page 75) tab.
Classification
• Select input band set : select the input Band set (page 31) to be classified;
• : open the Algorithm band weight (page 34) for the definition of band weights;
Algorithm
This tool allows for the selection of the classification algorithm and the optiona definition of thresholds.
• Threshold : it allows for the definition of a classification threshold (applied to all the spectral signatures); if T
– for Minimum Distance, pixels are unclassified if distance is greater than threshold value;
– for Maximum Likelihood, pixels are unclassified if probability is less than threshold value (max
100);
– for Spectral Angle Mapping, pixels are unclassified if spectral angle distance is greater than
threshold value (max 90);
• : open the Signature threshold (page 43) for the definition of signature thresholds;
Land Cover Signature Classification (page 139) is a classification that can be used as alternative or in combination
with the Algorithm (page 71) (see LCS threshold (page 39)). Pixels belonging to two or more different classes (or
macroclasses) are classified as Class overlap with raster value = -1000.
• Use LCS Algorithm only overlap: if LCS is checked, the Land Cover Signature Classification
is used; if Algorithm is checked, the selected Algorithm (page 71) is used for unclassified pixels of the Land
Cover Signature Classification; if only overlap is checked, the selected Algorithm (page 71) is used only for
class overlapping pixels of the Land Cover Signature Classification; unclassified pixels of the Land Cover
Signature Classification are left unclassified;
Classification output
Classification output allows for the classification of the Band set (page 31) according to the parameters defined in
Algorithm (page 71).
In addition, a previously saved classification style (QGIS .qml file) can be loaded and used for classification style.
Classification raster is a file .tif (a QGIS style file .qml is saved along with the classification); also other
outputs can be optionally calculated. Outputs are loaded in QGIS after the calculation.
• Load qml : select a .qml file overriding the colors defined for C ID or MC ID;
• : reset style to default (i.e. use the colors defined for C ID or MC ID);
• Apply mask : if checked, a shapefile can be selected for masking the classification output (i.e. the
area outside the shapefile is not classified);
• Create vector : if checked, in addition to the classification raster, a classification shapefile is saved
in the same directory and with the same name as the Classification output; conversion to vector can also be
performed at a later time (see Classification to vector (page 81));
• Classification report : if checked, a report about the land cover classification is calculated and
saved as a .csv file in the same directory and with the same name (with the suffix _report) as the Classi-
fication output; report can also be performed at a later time (see Classification report (page 81));
• Save algorithm files : if checked, the Algorithm raster (page 140) is saved, in addition
to the classification raster, in the same directory as the Classification output; a raster for each spectral
signature used as input (with the suffix _sig_MC ID_C ID) and a general algorithm raster (with the
suffix _alg_raster) are created;
• RUN : choose the output destination and start the image classification;
Clustering
This tab allows for the Clustering (page 146) of a Band set (page 31). In particular, K-means (page 146) and
ISODATA (page 146) methods are available.
A report .txt is saved along with the classification, containing the class spectral signature and the spectral distance
thereof.
• Select input band set : select the input Band set (page 31);
• Method K-means ISODATA: select the clustering method K-means (page 146) or ISODATA
(page 146);
• Distance threshold : if checked, for K-means: iteration is terminated if distance is lower than
threshold; for ISODATA: signatures are merged if distance is greater than threshold;
• Max number of iterations : maximum number of iterations if Distance threshold is not reached;
• ISODATA max standard deviation : maximum standard deviation considered for splitting a class, for
ISODATA algorithm only;
• ISODATA minimum class size in pixels : desired minimum class size in pixels, for ISODATA algorithm
only;
• Use value as NoData : if checked, set the value of NoData pixels, ignored during the calculation;
Seed signatures
• Seed signatures from band values Use Signature list as seed signatures Use random seed
signatures: select one options for seed signatures that start the iteration; the option Seed signatures from
band values divides the spectral space of the Band set (page 31) to get spectral signatures; the option Use
Signature list as seed signatures uses the spectral signatures checked in ROI & Signature list (page 24);
the option Use random seed signatures randomly selects the spectral signatures of pixels in the Band set
(page 31);
• Distance algorithm Minimum Distance Spectral Angle Mapping: select Minimum Distance
(page 135) or * Spectral Angle Mapping (page 137) for spectral distance calculation;
• Save resulting signatures to Signature list: if checked, save the resulting spectral signatures in the ROI
& Signature list (page 24);
PCA
This tab allows for the PCA (Principal Component Analysis (page 144)) of bands loaded in the Band set.
A report .txt is saved along with the PCA bands, containing the covariance matrix, correlation matrix, eigen
vectors, and eigen values.
• Select input band set : select the input Band set (page 31);
• Number of components : if checked, set the number of calculated components; if unchecked, all
the components are calculated;
• Use value as NoData : if checked, set the value of NoData pixels, ignored during the calculation;
• RUN : select an output directory and start the calculation process; principal components are calculated
and saved as raster files; also, the details about the PCA are displayed in the tab Output and saved in a .txt
file in the output directory;
Random Forest
This tool allows for the Random Forest (page 141) classification, based on a Band set (page 31) and Training input
(page 23).
ESA SNAP is required. The ESA SNAP GPT executable must be defined in External programs (page 112) settings.
• Select input band set : select the input Band set (page 31) to be classified;
• Number of training samples : set the number of training data (pixels) randomly used to traing the
model; it should be set lower than total training input pixels;
• Number of trees : set the number of decision trees; a higher number allows for more accurate models,
but it also increases the calculation time;
• Evaluate feature power set Min Max : if checked, evaluate the power set of input features
(e.g. Gini decrease), according to the contribution thereof to the model; Min and Max are used as thresholds
for power sets; it can increase the calculation time;
• Load classifier : open a previously saved classifier; if loaded, the input Band set (page 31) is
directly classified using this classifier;
Spectral distance
This tab allows for calculating the spectral distance between every corresponding pixel of two band sets. The
output is a raster containing the spectral distance of each pixel. Optionally, a threshold can be defined for creating
a binary raster (0-1) of values below and above the threshold.
• Select first input band set : select the first input Band set (page 31);
• Select second input band set : select the second input Band set (page 31);
• Distance algorithm Minimum Distance Spectral Angle Mapping: select Minimum Distance
(page 135) or * Spectral Angle Mapping (page 137) for spectral distance calculation;
• Distance threshold : if checked, a binary raster of values below and above the threshold is created;
3.4.6 Postprocessing
The tab Postprocessing provides several functions that can be applied to the classification output.
Accuracy
This tab allows for the validation of a classification (read Accuracy Assessment (page 143) ).
Classification is compared to a reference raster or reference vector (which is automatically converted to raster). If
a vector is selected as reference, it is possible to choose a field describing class values.
Several statistics are calculated such as overall accuracy, user’s accuracy, producer’s accuracy, and Kappa hat. In
particular, these statistics are calculated according to the area based error matrix where each element represents
the estimated area proportion of each class. This allows for estimating the unbiased user’s accuracy and producer’s
accuracy, the unbiased area of classes according to reference data, and the standard error of area estimates.
The output is an error raster that is a .tif file showing the errors in the map, where pixel values represent
the categories of comparison (i.e. combinations identified by the ErrorMatrixCode in the error matrix) be-
tween the classification and reference. Also, a text file containing the error matrix (i.e. a .csv file separated by
tab) is created with the same name defined for the .tif file.
Accuracy assessment
• Select the classification to assess : select a classification raster (already loaded in QGIS);
• Use value as NoData : if checked, pixels having NoData value are excluded from the calculation;
• RUN : choose the output destination and start the calculation; the error matrix is displayed in the tab
Output and the error raster is loaded in QGIS;
Classification dilation
This tab allows for dilating the border of a class patch, defining the class values to be dilated and the number of
pixels from the border. It is useful for classification refinement.
Classification dilation
• Class values : set the class values to be dilated; class values must be separated by , and - can be used
to define a range of values (e.g. 1, 3-5, 8 will select classes 1, 3, 4, 5, 8); if the text is red then the
expression contains errors;
Classification erosion
This tab allows for removing the border of a class patch (erosion), defining the class values to be eroded and the
number of pixels from the border. It is useful for classification refinement.
Classification erosion
• Class values : set the class values to be eroded; class values must be separated by , and - can be used
to define a range of values (e.g. 1, 3-5, 8 will select classes 1, 3, 4, 5, 8); if the text is red then the
expression contains errors;
Classification report
This tab allows for the calculation of class statistics such as number of pixels, percentage and area (area unit is
defined from the image itself).
Classification report
• Use value as NoData : if checked, NoData value will be excluded from the report;
• RUN : choose the output destination and start the calculation; the report is saved in a text file and
displayed in the tab Output;
Classification to vector
This tab allows for the conversion of a classification raster to vector shapefile.
Classification to vector
Symbology
• Use code from Signature list : if checked, color and class information are defined from ROI & Signature list (p
Classification sieve
This tab allows for the replacement of isolated pixel values with the value of the largest neighbour patch (based
on GDAL Sieve ). It is useful for removing small patches from a classification.
Classification sieve
• Size threshold : size of the patch to be replaced (in pixel unit); all patches smaller than the selected
number of pixels will be replaced by the value of the largest neighbour patch;
• Pixel connection : select the type of pixel connection:
– 4: in a 3x3 window, diagonal pixels are not considered connected;
– 8: in a 3x3 window, diagonal pixels are considered connected;
Class signature
This tab allows for the calculation of the mean spectral signature of each class in a classification using a Band set
(page 31).
• Select the classification : select a classification raster (already loaded in QGIS);
• Select input band set : select the input Band set (page 31) for spectral signature calculation;
• Save resulting signatures to Signature list: if checked, save the resulting spectral signatures to ROI &
Signature list (page 24);
Cross classification
This tab allows for the calculation of a cross classification raster and matrix. Classification is compared to a
reference raster or reference vector (which is automatically converted to raster). This is useful for calculating the
area for every combination between reference classes and classification values. If a vector is selected as reference,
it is possible to choose a field describing class values.
The output is a cross raster that is a .tif file where pixel values represent the categories of comparison
(i.e. combinations identified by the CrossMatrixCode) between the classification and reference. Also, a text
file containing the cross matrix (i.e. a .csv file separated by tab) is created with the same name defined for the
.tif file.
Input
• Use value as NoData : if checked, NoData value will be excluded from the calculation;
• Select the reference vector or raster : select a raster or a vector (already loaded in QGIS), used as
reference layer;
• RUN : choose the output destination and start the calculation; the cross matrix is displayed in the tab
Output and the cross raster is loaded in QGIS;
Edit raster
This tab allows for the direct editing of pixel values in a raster. Only pixels beneath ROI polygons or vector
polygons are edited.
WARNING: the input raster is directly edited; it is recommended to create a backup copy of the
input raster before using this tool in order to prevent data loss.
This tool can rapidly edit large rasters, especially when editing polygons are small, because pixel values are edited
directly. In addition, the SCP Edit Toolbar (page 120) is available for easing the raster editing using multiple
values.
Edit raster
• Edit values using ROI polygons: if checked, raster is edited using temporary ROI polygons in the map;
• Edit values using a vector : if checked, raster is edited using all the polygons of selected vector;
Edit options
• Use the value field of the vector : if checked, raster is edited using the selected vector (in Edit
values using a vector) and the polygon values of selected vector field;
• Use constant value : if checked, raster is edited using the selected constant value;
• Use expression : if checked, raster is edited according to the entered expression; the expression
must contain one or more where; accepted variable are raster representing the input raster value and
vector representing the vector value if selected; the following example expression where(raster ==
1, 2, raster) is already entered, which sets 2 where raster equals 1, and leaves unchanged the
values where raster is not equal to 1;
• : undo the last raster edit (available only when using ROI polygons);
The tab Land cover change allows for the comparison between two classifications in order to assess land
cover changes.
Output is a land cover change raster (i.e. a file .tif showing the changes in the map, where each
pixel represents a category of comparison (i.e. combinations) between the two classifications, which is the
ChangeCode in the land cover change statistics) and a text file containing the land cover change statistics (i.e. a
file .csv separated by tab, with the same name defined for the .tif file).
Input
• Select the reference classification : select a reference classification raster (already loaded in QGIS);
• Select the new classification : select a new classification raster (already loaded in QGIS), to be com-
pared with the reference classification;
• Report unchanged pixels: if checked, report also unchanged pixels (having the same value in both
classifications);
• RUN : choose the output destination and start the calculation; the land cover change statistics are
displayed in the tab Output (and saved in a text file) and the land cover change raster is loaded
in QGIS;
Reclassification
This tab allows for the reclassification (i.e. assigning a new class code to raster pixels). In particular, it eases the
conversion from C ID to MC ID values.
Reclassification
Values
• calculate C ID to MC ID values: if checked, the reclassification table is filled according to the ROI &
• Calculate unique values : calculate unique values in the classification and fill the reclassification table;
• Incremental new values : calculate a new incremental value (e.g. 1, 2, 3, . . . ) for every old value and
fill the reclassification table;
Symbology
• Use code from Signature list : if checked, color and class information are defined from ROI & Signature list (p
• RUN : choose the output destination and start the calculation; reclassified raster is loaded in QGIS;
This tab allows for calculating the statistics related to an input raster for every unique value of a reference vector
or raster. For example, if a vector file contains a numeric field zone, this tool allows for calculating the mean
value (for each zone) of input raster pixels.
• Use value as NoData : if checked, pixels having NoData value are excluded from the calculation;
• Select the reference vector or raster : select a raster or a vector (already loaded in QGIS), used as
reference layer for zones;
Statistic
• Select a statistics : select a statistic; available options are Sum, Max, Min, Count, Mean, Median,
Percentile, StandardDeviation;
The Band calc allows for the raster calculation for bands (i.e. calculation of pixel values) using NumPy
functions . Band calc can also work with multuple expression lines to perform multiple calculation at a time;
several iteration functions are available for Band sets allowing to perform multitemporal calculations.
Raster bands must be already loaded in QGIS. Input rasters must be in the same projection.
In addition, a tool for decision rules is included.
Band list
• Band list: table containing a list of single band rasters (already loaded in QGIS);
Expression
Enter a mathematical expression for raster bands. In particular, NumPy functions can be used with the prefix np.
(e.g. np.log10(raster1) ). For a list of NumPy functions see the NumPy page .
The expression can work with both Variable and Band name (between double quotes). Double click on any item
in the Band list (page 91) for adding the name thereof to the expression.
If text color in the Expression is green, then the syntax is correct; if text color is red, then the syntax is incorrect
and it is not possible to execute the calculation.
It is possible to enter multiple expressions separated by newlines such as the following example:
"raster1" + "raster2"
"raster3" - "raster4"
The above example calculates two new rasters in the output directory with the suffix _1 (e.g. calc_raster_1
) for the first expression and _2 (e.g. calc_raster_2 ) for the second expression. Also, it is possible to define
the output name using the symbol @ followed by the name, such as the following example:
• Input NoData as value: if checked, input NoData pixels are replaced by NoData value and accounted
in calculation;
• Use value as NoData : if checked, selected value is used as NoData during calculation;
• Extent: if the following options are unchecked, the output raster extent will include the extents of all input rasters;
– Intersection: if checked, the extent of output raster equals the intersection of input raster
extents (i.e. minimum extent);
– Same as : if checked, the extent of output raster equals the extent of “Map extent” (the
extent of the map currently displayed) or a selected layer;
• Align: if checked, and Same as is checked selecting a raster, the calculation is performed using
the same extent and pixel alignment of selected raster;
Input variables
The following variables related to the Band set (page 31) are available (please note that the character " is always
required):
• bandset#b BAND_NUMBER: bands in the active Band set can be referenced directly; the following example
refers to band 1 of the active Band set:
"bandset#b1"
• bandset BANDSET_NUMBER b BAND_NUMBER: bands in the Band set can be referenced directly; the
following example refers to band 1 of the Band set 1:
"bandset1b1"
• bandset{ BANDSET_NUMBER_LIST }b BAND_NUMBER: list of all the bands X of all the Band sets
matching a list between curly brackets (e.g. {1,2,3}) or range of Band sets separated by colon (e.g. {1:3})
or a list of ranges of Band sets (e.g. {1:3, 5:8}), which is equivalent to [bandset1bX, bandset2bX,
..., bandsetNbX]; this variable can be used in expressions that accept band lists such as the sum:
sum("bandset{1,3}b1")
• bandset#b*: list of all the bands of active Band set, which is equivalent to [bandset#b1,
bandset#b2, ..., bandset#bX]; to be used in expressions that accept band lists such as the max-
imum value:
max("bandset#b*")
• bandset BANDSET_NUMBER b*: list of all the bands of Band set N, which is equivalent to
[bandsetNb1, bandsetNb2, ..., bandsetNbX]; to be used in expressions that accept band
lists such as the minimum value:
min("bandset1b*")
• bandset*b BAND_NUMBER: list of all the bands X of all the Band sets, which is equivalent to
[bandset1bX, bandset2bX, ..., bandsetNbX]; to be used in expressions that accept band
lists such as the mean value:
mean("bandset*b1")
• bandset{ DATE }b BAND_NUMBER: list of all the bands X of all the Band sets matching a list of dates
(format yyyy-mm-dd) between curly brackets (e.g. {2019-01-01,2019-07-31}) or range of dates separated
by colon (e.g. {2019-01-01:2019-07-31}) or a list of ranges of dates (e.g. {2019-01-01:2019-01-31, 2019-
04-01:2019-07-31}), which is equivalent to [bandset2bX, bandset5bX, ..., bandsetNbX];
Band sets dates are defined in Band set definition (page 32); this variable can be used in expressions that
accept band lists such as the median value:
median("bandset{2019-01-01,2019-07-31}b1")
• “#BLUE#”: the band with the center wavelength closest to 0.475 𝜇𝑚;
• “#GREEN#”: the band with the center wavelength closest to 0.56 𝜇𝑚;
• “#RED#”: the band with the center wavelength closest to 0.65 𝜇𝑚;
• “#NIR#”: the band with the center wavelength closest to 0.85 𝜇𝑚; for example:
Output variables
Output names can be defined in the expression line entering the symbol @ followed by the name, such as the
following example:
"raster1" * 2 @first_calculation
If the output name is defined in the expression, an output directory will be selected after clicking the button RUN.
WARNING: It is recommended to avoid the use of characters # and @ in the output raster name.
It is possible to set the output path directly by defining the output name with this structure @path@name, such
as:
"raster1" * 2 @/home/user@first_calculation
The output name of calculation can be used as input variable for the following calculations; for example:
"raster1" * 2 @first_calculation
"first_calculation" + 5 @temp@second_calculation
"second_calculation" - "raster1" @/home/user@third_calculation
It is possible to add the calculation result to a Band set using the symbol % followed by the Band set number
such as:
"raster1" @first_calculation%1
It is possible to add the calculation result to the active Band set using the symbol % followed by the symbol #
such as:
"raster1" @first_calculation%#
If the output name is defined with the extension .vrt, the output will be a virtual raster referencing the single
.tif files calculated by parallel processing. For instance, the following expression will calculate the raster with
parallel processing and the output will be a .vrt:
"raster1" @first_calculation.vrt
Virtual file is useful to reduce calculation time of parallel processing by avoiding the writing of the entire output
raster.
WARNING: a .vrt virtual file is a text file containing the reference to existing .tif files. Moving
the .vrt virtual file or the .tif files in different directories can make the virtual file unusable.
Buttons
• : open a text file (.txt) containing custom expressions to be listed in Functions; the text file must
contain an expression for each line; each line must be in the form expression_name; expression
(separated by ;) where the expression_name is the expression name that is displayed in the Functions;
if you open an empty text file, the default values are restored; following an example of text content:
Functions
Conditional
• where: conditional expression according to the syntax where( condition , value if true,
value if false) (e.g. where("raster1" == 1, 2, "raster1"));
Logical
• AND: AND;
• OR: OR;
• XOR: XOR;
• NOT: NOT;
Statistics
A band list between square brackets or Input variables (page 92) are required arguments.
• max: maximum; for instance:
min(["bandset#b*"])
mean("bandset*b1")
median("bandset{2019-01-01,2019-07-31}b1")
• percentile: percentile calculation; the expression must have this structure percentile([band_list],
percentile_value); for instance, the following expression calculates the 10th percentile of active
band set:
percentile("bandset#b*", 10)
Operations
• sin: sine; for instance:
sin("raster1")
• cos: cosine;
• tan: tangent;
• asin: inverse sine;
• acos: inverse cosine;
• atan: inverse tangent;
• exp: natural exponential;
• ln: natural logarithm;
• log: base 10 logarithm;
Indices
• NDVI: if selected, the NDVI calculation is entered in the Expression:
( "#NIR#" - "#RED#" ) / ( "#NIR#" + "#RED#" ) @ NDVI
Variables
• nodata: NoData value of raster (e.g. nodata("raster1")); it can be used as value in the expression:
where("raster1" == nodata("raster1"), 0, "raster1")
• forbandsets: it allows for the iteration over Band sets defined between square brackets; a range of Band sets
separated by colon (e.g. forbandsets[1:3]) or a list separated by commas (e.g. forbandsets[1,
2,3]; forbandsets must be entered in the first line of expressions (not compatible with forbsdates
or forbandsinbandset), for instance:
forbandsets[1:3]
"bandset#b1" @#BANDSET#
During the iteration, the Input variables (page 92) related to the active Band set are replaced by the iterator,
therefore the expression:
forbandsets[1:3]
"bandset#b1" @#BANDSET#
is equivalent to:
"bandset1b1" @calc1
"bandset2b1" @calc2
"bandset3b1" @calc3
It is possible to enter a string (or a list of strings separated by comma ,) after the square bracket ] to filter Band
sets by the name of first band in Band set; for instance, the following expression will iterate the first 3 Band sets
selecting only the Band set whose name of the first band includes RT:
forbandsets[1,2,3]RT
• forbsdates: it allows for the iteration over Band sets dates (format yyyy-mm-dd, defined in Band set defini-
tion (page 32)) between square brackets; forbsdates must be entered in the first line of expressions (not
compatible with forbandsets or forbandsinbandset) such as the following examples:
Iterating over a range of dates:
forbsdates[2020-01-01:2020-07-31]
"bandset#b1" @#BANDSET#
forbsdates[2020-02-01,2020-03-11,2020-04-21]
"bandset#b1" @#BANDSET#
During the iteration, the Input variables (page 92) related to the active Band set are replaced by the iterator.
It is possible to enter a string (or a list of strings separated by comma ,) after the square bracket ] to filter Band
sets by the name of first band in Band set; for instance, the following expression will iterate the Band sets in the
range selecting only the Band set whose name of the first band includes RT:
forbsdates[2020-01-01:2020-07-31]RT
• forbandsinbandset: it allows for the iteration over bands in a Band set or a range of Band sets between
square brackets; forbandsinbandset must be entered in the first line of expressions (not compatible
with forbandsets or forbsdates); the variable #BAND# must be used to refer to the iterated band
such as the following examples:
Iterating over the bands of the first Band set (note the " in the expression):
forbandsinbandset[1]
where("#BAND#" > 1, 1, 2) @#BAND#)
forbandsinbandset[1:3]
where("#BAND#" > 1, 1, 2) @#BAND#)
• #BAND#:
• !function!: it allows for using a Batch (page 99) function inside Band calc; !function! must be followed by
a Batch command to execute a function using raster names (also using Output variables (page 93)) in the
command; the special variable !directory! can be used in the Batch command to refer to the output
directory (defined after clicking the button RUN); the special variable !file! can be used in the Batch
command to refer to the relative output file name (defined with @ after the command) respectively ; it
is required to use the section character $ instead of apostrophe ' and !! instead of ; for functions; for
instance:
• #BANDSET#: the name of the first band in the Band set (page 31);
• #DATE#: the current date and time (e.g. 20161110_113846527764);
• @: charcter @;
Decision rules
Decision rules allows for the calculation of an output raster based on rules. Rules are conditional statements based
on other rasters; if the Rule is true, the corresponding Value is assigned to the output pixel.
Rules are verified from the first to the last row in the table; if the first Rule is false, the next Rule is verified for that
pixel, until the last rule. If multiple rules are true for a certain pixel, the value of the first Rule is assigned to that
pixel. The NoData value is assigned to those pixels where no Rule is true.
Output raster
The output raster is a .tif file, with the same spatial resolution and projection of input rasters; if input rasters have
different spatial resolutions, then the highest resolution (i.e. minimum pixel size) is used for output raster.
WARNING: if multiple lines are entered in Expression or output name are defined in line, a directory
selection instead of a file selection is prompted for output.
• Float32 : option to select output raster type as Float32 (default), Int32, Int16, UInt32, UInt16, Byte;
• Set scale : if checked, set a scale in output raster (and output values are divided by scale);
• Set offset : if checked, set an offset in output raster (and offset is subtracted to output values);
• RUN : if Expression is active and text is green, choose the output destination and start the cal-
culation based on Expression; if Decision rules is active and text is green, choose the output
destination and start the calculation based on Decision rules;
3.4.8 Batch
This tab allows for the automatic execution (batch) of several SCP functions using a scripting interface.
Batch
Enter a batch expression; each function must be in a new line. Functions have the following structure:
Each function has options, identified by a name, with the following structure:
Options must be separated by the character ; . Each function option represents an option in the corresponding
interface of SCP; option arguments of type text must be between the character ' ; in case of checkboxes, the
value 1 represents checked, while the value 0 represents unchecked. A new line beginning with # can be used for
commenting.
According to the function, some of the options are mandatory while other options can be omitted from the expres-
sion. Option names that contain path require the full path to a file.
Some options require multiple arguments such as lists; lists must be separated by , .
Band set
• add_new_bandset: add a new empty band set; optionally it is possible to specify the position of the added
band set (e.g. band_set : 1):
add_new_bandset;band_set : 1
• add_raster: add a raster to QGIS, and optionally to an existing band set defining the wavelength:
• create_bandset: create a Band set assigning bands; it is possible to enter the satellite name
as center_wavelength option (satellite name as displayed in Wavelength quick settings e.g.
center_wavelength : 'sentinel-2' or center_wavelength : 'landsat 8') to
set the corresponding band wavelength center automatically for all the bands; raster_path_list op-
tion accepts a directory and a file name filter separated by comma (e.g. raster_path_list : '/
home/user, tif') to load all the raster inside the directory:
create_bandset;raster_path_list : '';center_wavelength : '';wavelength_unit :
˓→1;multiplicative_factor : '';additive_factor : '';date : ''
• remove_band_from_bandset: remove a list band from band set by the number thereof (e.g.
remove_band_from_bandset;band_set : 1;band_list : '1, 2'):
remove_band_from_bandset;band_set : 1;band_list : ''
• select_bandset: select a band set (active band set) by the number thereof:
select_bandset;band_set : 1
Band calculation
˓→1;offset_value : 0;band_set : 1
Preprocessing
˓→: 1
˓→dir : ''
˓→1;output_dir : '';band_set : 1
• Stack raster bands (page 67): stack rasters into a single file:
Band preprocessing
˓→raster_path : ''
˓→classification_path : ''
Post preprocessing
classification_to_vector;input_raster_path : '';use_signature_list_code : 0;
˓→code_field : 'C_ID';dissolve_output : 0;output_vector_path : ''
˓→: ''
• Zonal stat raster (page 89): calculate zonal statistics related to an input raster:
˓→output_raster_path : ''
Variables
Working directory
• !working_dir!: set a working directory (argument is the path to a directory):
!working_dir!;''
!working_dir!; '/home/user/Desktop/temp/'
add_raster;input_raster_path : '!working_dir!/raster1.tif';input_raster_name :
˓→'raster1.tif'
Iterate directories
• !for_directory_in!: function to iterate commands for all the directories in a directory path (e.g. !
for_directory_in!;'directoryPath'):
!for_directory_in!;''
The variables !directory_name! can be used to reference the current directory name and !directory! for the current
directory path.
The iterated commands must terminate with a line !end_for_directory!; for instance the following script
performs Sentinel-2 conversion of all the subdirectories of /home/user/:
!for_directory_in!;'/home/user/'
sentinel2_conversion;input_dir : '!directory!';create_bandset : 1;output_dir : '!
˓→directory!_converted'
!end_for_directory!
Optionally a filter (i.e. only directories matching the filter will be processed) can be defined for the subdirectory
level; the following examples stop the iteration at level 2:
!for_directory_in!;'/home/user/';2
A name filter can be set; for instance the following script creates a band set for every subdirectory having rt in
the name:
!for_directory_in!;'/home/user/';'rt'
add_new_bandset;
create_bandset;raster_path_list : '!directory!, tif';center_wavelength : 'landsat 8
˓→';
!end_for_directory!
!for_directory_in!;'/home/user/';3;'name1|name2')
It is possible to filter by date if the directory name ends with a date (format yyyy-mm-dd) using a range separated
by colon:
!for_directory_in!;'/home/user/';2020-01-01:2020-03-31
sentinel2_conversion;input_dir : '!directory!';create_bandset : 1;output_dir : '!
˓→directory!_converted'
!end_for_directory!
!directory!
!directory_name!
• !end_for_directory!: ending line to be used with !for_directory_in! to close the iterative com-
mands:
!end_for_directory!
Iterate files
• !for_file_in!: function to iterate commands for all the files in a directory path (e.g. !for_file_in!;
'directoryPath'):
!for_file_in!;''
The variable !file! can be used to reference the current file path.
The iterated commands must terminate with a line !end_for_file!, for instance:
!for_file_in!;'/home/user/'
create_bandset;raster_path_list : '!file!';
!end_for_file!
Optionally a filter (i.e. only directories matching the filter will be processed) can be defined for the subdirectory
level; the following examples stop the iteration at level 0:
!for_file_in!;'/home/user/';0
!for_file_in!;'/home/user/';0;'tif'
!for_file_in!;'/home/user/';0;'jp2|tif'
It is possible to filter by date if the file name ends with a date (format yyyy-mm-dd) using a range:
!for_file_in!;'/home/user/';2020-01-01:2020-03-31
Also !file_directory! can be used to refer to the parent directory of the file.
• !file!: variable to be used with !for_file_in! to reference current file path:
!file!
• !file_directory!: variable to be used with !for_file_in! to reference the directory containing current
file path:
!file_directory!
• !end_for_file!: ending line to be used with !for_file_in! to close the iterative commands:
!end_for_file!
!for_band_set!;''
It is possible to enter a list of numbers (separated by commas), a range separated by colon, or a list of ranges, for
instance:
!start_for_band_set!;'2:3'
add_new_bandset;band_set : !bandset_number!
!end_for_band_set!
!number! = 20
!path_to_file! = /home/file.tif
add_new_bandset;band_set : !number!
band_combination;band_set : 1;output_raster_path : '!path_to_file!'
Temporary output
• !temp_raster_1!: create a temporary raster path (saved in the temporary directory) to be used with other
functions:
!temp_raster_1!
!temp_raster_1! can be used as temporary output for other functions such as:
Additional temporary output can be entered manually, replacing 1 with any number or string (e.g. !temp_raster_2!
or !temp_raster_calculation! ). The actual raster name will be the string between ! , for instance temp_raster_1 is
the raster name of variable !temp_raster_1!.
QGIS Processing
• qgis_processing: use a QGIS Processing algorithm entering a command (algorithm name) and parameters
(without brackets) as described in the QGIS console
˓→TYPE_PARAMETER':0
It is possible to use the Batch Variables (page 104) for setting input or output paths.
Run
• CHECK : check if expression syntax is correct; if expression is correct then the text color is green, red
otherwise; in case of error, a label is displayed with a brief description of the error;
• RUN : check if expression syntax is correct and start the batch processes;
3.4.9 Settings
Debug
Debugging utilities for the creation of a Log file (i.e. recording of SCP activities for reporting issues) and testing
SCP dependencies.
http://www.youtube.com/watch?v=So04gNzDC9Y
If you found a plugin error, please read How can I report an error? (page 204) .
Log file
• Records events in a log file : if checked, start recording events in a Log file;
Test
• Test dependencies : test SCP dependencies (GDAL, GDAL subprocess, NumPy, SciPy, Matplotlib,
Internet connection); results are displayed in a window;
Interface
Set the names of fields in the Training input (page 23) . Changing field names should usually be avoided.
ROI style
Change ROI color and transparency for a better visualization of temporary ROIs on the map.
Set the variable name used in expressions of the Reclassification (page 88) and Edit raster (page 85) .
Set the temporary group name in QGIS Layers used for temporary layers .
• Create RGB composite of band set when a project is loaded : if checked, a RGB composite of
active band set is added to the map when a project is loaded;
Dock
• Download news on startup : if checked, news about the SCP and related services are downloaded
on startup and displayed in Dock;
Processing
System
• Available RAM (MB) : set the available RAM (in MB) that is used during the processes in order
to improve the SCP performance; this value should be half of the system RAM (e.g. 1024MB if system has
2GB of RAM); in case of errors, set a value lower than 512MB;
• CPU threads : set the number of threads available for processing; it is recommended to set a
value lower than the maximum number of system threads (e.g. if the system has 4 available threads set
value 3);
http://www.youtube.com/watch?v=KDI3ud9L6UY
Classification process
• Play sound when finished : if checked, play a sound when the classification process is completed;
• Use virtual rasters for temp files : if checked, create virtual rasters for certain temporary files,
instead of creating real rasters; it is useful for reducing disk space usage during calculations;
• Enable writing verification : if checked, the output raster is compared to output calculation in
order to avoid writing errors; it can slow the process;
• SMTP server : the SMTP server to login for sending a notification email when all the SCP
processes are finished (a notification is sent also in case of error, but not in case of crash);
• Send email of completed process to : a list of addresses (separated by comma) to send the notifi-
cation email to;
External programs
• ESA SNAP GPT executable : enter the path to the SNAP gpt executable; usually
C:/snap/bin/gpt.exe for Windows, /Applications/snap/bin/gpt for Mac, and /usr/
local/snap/bin/gpt for Linux; for the installation of ESA SNAP, please visit this page http:
//step.esa.int/main/download/snap-download/ ;
• Python executable path : enter the path to the Python 3 executable (e.g. /usr/local/
bin/python3); should be used only in case of issues with multiprocessing;
• GDAL installation directory : enter the path to the GDAL directory containing tools such
as gdal_translate and gdalwarp (e.g. /usr/bin); should be used only in case of issues with GDAL pro-
cessing;
Temporary directory
• : select a new temporary directory where temporary files are saved during processing; the path to
the current temporary directory is displayed; default is a system temporary directory;
The window Spectral Signature Plot includes several functions for displaying spectral signature values as a func-
tion of wavelength (defined in the Band set (page 31)). Signatures can be added to the Spectral Signature
Plot through the SCP dock (page 20).
The window Spectral Signature Plot includes also some functions useful for the definition of value ranges used
by the Land Cover Signature Classification (page 139) (see LCS threshold (page 39)).
Overlapping signatures (belonging to different classes or macroclasses) are highlighted in orange in the table
Plot Signature list (page 113); the overlapping check is performed considering MC ID or C ID according to the
setting Use MC ID C ID in Algorithm (page 71). Overlapping signatures sharing the same ID are not
highlighted.
• Signature list:
– S: checkbox field; if checked, the spectral signature is displayed in the plot;
– MC ID: signature Macroclass ID;
– MC Name: signature Macroclass Name;
– C ID: signature Class ID;
– C Name: signature Class Name;
– Color [overlap MC_ID-C_ID]: signature color; also, the combination MC ID-C ID is displayed
in case of overlap with other signatures (see Land Cover Signature Classification (page 139));
– Min B X: minimum value of band X; this value can be edited;
– Max B X: maximum value of band X; this value can be edited;
• : add highlighted spectral signatures to ROI & Signature list (page 24);
• : calculate the spectral distances of spectral signatures displayed in the plot; distances are reported in
the tab Spectral distances (page 116);
Automatic thresholds
Set thresholds automatically for highlighted signatures in the table Plot Signature list (page 113); if no signature
is highlighted, then the threshold is applied to all the signatures.
• Min Max : set the threshold based on the minimum and maximum of each band;
• 𝜎* : set an automatic threshold calculated as (band value + (𝜎 * v)), where 𝜎 is the standard
deviation of each band and v is the defined value;
• From ROI : set the threshold using the temporary ROI pixel values, according to the following checkboxes:
• From pixel : set the threshold by clicking on a pixel, according to the following checkboxes:
Plot
Left click and hold inside the plot to move the view of the plot. Use the mouse wheel to zoom in and out the
view of the plot. Right click and hold inside the plot to zoom in a specific area of the plot. Legend inside the
plot can be moved using the mouse.
Plot commands:
• : save the plot image to file (available formats are .jpg, .png, and .pdf);
• : activate the cursor for interactively changing the value range of highlighted signatures in the plot;
click the plot to set the minimum or maximum value of a band (also for several signatures simultaneously);
cursor is deactivated when moving outside the plot area;
• Plot value range: if checked, plot the value range for each signature (semi-transparent area);
• Band lines: if checked, display a vertical line for each band (center wavelength);
Signature details
Display the details about spectral signatures (i.e. Wavelength, Values, and Standard deviation). In case of signa-
tures calculated from ROIs, the ROI size (number of pixels) is also displayed.
Spectral distances
Display spectral distances of signatures (see Plot Signature list (page 113)), which are useful for assessing ROI
separability (see Spectral Distance (page 141)).
The following spectral distances are calculated :
• Jeffries-Matusita Distance (page 141): range [0 = identical, 2 = different]; useful in particular for
Maximum Likelihood (page 136) classifications;
• Spectral Angle (page 142): range [0 = identical, 90 = different]; useful in particular for Spectral Angle
Mapping (page 137) classifications;
• Euclidean Distance (page 142): useful in particular for Minimum Distance (page 135) classifications;
• Bray-Curtis Similarity (page 142): range [0 = different, 100 = identical]; useful in general;
Values are displayed in red if signatures are particularly similar.
The window Scatter plot displays pixel values for two raster bands as points in the 2D space. Scatter plots are
useful for assessing ROI separability between two bands.
The functions are described in detail in the following paragraphs.
• Scatter list:
– S: checkbox field; if checked, the spectral signature is displayed in the plot;
– MC ID: signature Macroclass ID;
– MC Name: signature Macroclass Name;
– C ID: signature Class ID;
– C Name: signature Class Name;
– Color: color field; double click to select a color for the plot;
• Precision : use custom precision for calculation (precision should be selected according to pixel values):
– 4 = 10−4
– 3 = 10−3
– 2 = 10−2
– 1 = 10−1
– 0=1
– -1 = 10
– -2 = 102
– -3 = 103
• Calculate : calculate the scatter plot for the ROIs checked in the list;
• : add a temporary scatter plot to the list (as MC Name = tempScatter) and start the plot calcula-
tion of the last temporary ROI (see Working toolbar (page 27));
• : add a temporary scatter plot to the list (as MC Name = tempScatter) and start the plot calcula-
tion of pixels in current display extent;
• : add a temporary scatter plot to the list (as MC Name = tempScatter) and start the plot calcula-
tion of the entire image;
WARNING: Using a precision value that is too high can result in slow calculation or failure.
Scatter raster
This tool allows for the drawing of selection polygons inside the scatter plot; these selection polygons are used
for creating a Scatter raster that is a temporary raster classified according to the intersection of scatter plots and
drawn polygons.
Pixels of the active band set are classified, according to scatter plot bands, if pixel values are in the range of
intersection between scatter plots and selection polygons (polygons should not overlap). The value assigned to
the Scatter raster pixels is the sequential number of selection polygon; also the raster color is derived from the
selection polygon.
After the creation of a new Scatter raster, old rasters are placed in QGIS Layers inside a layer group named
Class_temp_group (custom name can be defined in Temporary group name (page 110)) and are deleted when
the QGIS session is closed.
• : activate the cursor for interactively drawing a polygon in the plot; left click on the plot to define the
vertices and right click to define the last vertex closing the polygon;
• color: select the color of polygon (which is used also in the Scatter raster);
• : calculate the spectral signature of the Scatter raster (considering all the classified pixels) using the
active band set, and save the signature to the ROI & Signature list (page 24);
• Extent : extent of the Scatter raster; available options are:
– Same as display: extent is the same as map display;
– Same as image: extent is the same as the whole image;
Plot
Left click and hold inside the plot to move the view of the plot. Use the mouse wheel to zoom in and out the
view of the plot. Right click and hold inside the plot to zoom in a specific area of the plot.
• Colormap : select a colormap that is applied to highlighted scatter plots in the list when is
clicked; if no scatter plot is highlighted then the colormap is applied to all the scatter plots;
• : save the plot image to file (available formats are .jpg, .png, and .pdf);
• x y: display x y coordinates of mouse cursor inside the plot;
The SCP Edit Toolbar allows for the direct editing of pixel values in the input raster defined in Edit raster (page 85)
using ROI polygons. Only pixels beneath ROI polygons are edited.
• : open the tool SCP Edit Toolbar for selecting the input raster;
• : undo the last raster edit (available only when using ROI polygons);
= Input text
= List
= Input number
= Optional
= Slider
= Table
123
Semi-Automatic Classification Plugin Documentation, Release 7.5.5.1
This chapter provides basic definitions about GIS and remote sensing. For other useful resources see Free and
valuable resources about remote sensing and GIS (page 207).
There are several definitions of GIS (Geographic Information Systems), which is not simply a program. In general,
GIS are systems that allow for the use of geographic information (data have spatial coordinates). In particular,
GIS allow for the view, query, calculation and analysis of spatial data, which are mainly distinguished in raster
or vector data structures. Vector is made of objects that can be points, lines or polygons, and each object can
have one ore more attribute values; a raster is a grid (or image) where each cell has an attribute value (Fisher and
Unwin, 2005). Several GIS applications use raster images that are derived from remote sensing.
A general definition of Remote Sensing is “the science and technology by which the characteristics of objects of
interest can be identified, measured or analyzed the characteristics without direct contact” (JARS, 1993).
Usually, remote sensing is the measurement of the energy that is emanated from the Earth’s surface. If the source
of the measured energy is the sun, then it is called passive remote sensing, and the result of this measurement can
be a digital image (Richards and Jia, 2006). If the measured energy is not emitted by the Sun but from the sensor
platform then it is defined as active remote sensing, such as radar sensors which work in the microwave range
(Richards and Jia, 2006).
The electromagnetic spectrum is “the system that classifies, according to wavelength, all energy (from short
cosmic to long radio) that moves, harmonically, at the constant velocity of light” (NASA, 2013). Passive sensors
measure energy from the optical regions of the electromagnetic spectrum: visible, near infrared (i.e. IR), short-
wave IR, and thermal IR (see Figure Electromagnetic-Spectrum (page 125)).
Fig. 1: Electromagnetic-Spectrum
by Victor Blacus (SVG version of File:Electromagnetic-Spectrum.png)
4.1. Basic Definitions
[CC-BY-SA-3.0 (http://creativecommons.org/licenses/by-sa/3.0)] 125
via Wikimedia Commons
http://commons.wikimedia.org/wiki/File%3AElectromagnetic-Spectrum.svg
Semi-Automatic Classification Plugin Documentation, Release 7.5.5.1
The interaction between solar energy and materials depends on the wavelength; solar energy goes from the Sun to
the Earth and then to the sensor. Along this path, solar energy is (NASA, 2013):
• Transmitted - The energy passes through with a change in velocity as determined by the index of refraction
for the two media in question.
• Absorbed - The energy is given up to the object through electron or molecular reactions.
• Reflected - The energy is returned unchanged with the angle of incidence equal to the angle of reflec-
tion. Reflectance is the ratio of reflected energy to that incident on a body. The wavelength reflected (not
absorbed) determines the color of an object.
• Scattered - The direction of energy propagation is randomly changed. Rayleigh and Mie scatter are the two
most important types of scatter in the atmosphere.
• Emitted - Actually, the energy is first absorbed, then re-emitted, usually at longer wavelengths. The object
heats up.
4.1.3 Sensors
Sensors can be on board of airplanes or on board of satellites, measuring the electromagnetic radiation at specific
ranges (usually called bands). As a result, the measures are quantized and converted into a digital image, where
each picture elements (i.e. pixel) has a discrete value in units of Digital Number (DN) (NASA, 2013). The
resulting images have different characteristics (resolutions) depending on the sensor. There are several kinds of
resolutions:
• Spatial resolution, usually measured in pixel size, “is the resolving power of an instrument needed for the
discrimination of features and is based on detector size, focal length, and sensor altitude” (NASA, 2013);
spatial resolution is also referred to as geometric resolution or IFOV;
• Spectral resolution, is the number and location in the electromagnetic spectrum (defined by two wave-
lengths) of the spectral bands (NASA, 2013) in multispectral sensors, for each band corresponds an image;
• Radiometric resolution, usually measured in bits (binary digits), is the range of available brightness values,
which in the image correspond to the maximum range of DNs; for example an image with 8 bit resolution
has 256 levels of brightness (Richards and Jia, 2006);
• For satellites sensors, there is also the temporal resolution, which is the time required for revisiting the
same area of the Earth (NASA, 2013).
Sensors measure the radiance, which corresponds to the brightness in a given direction toward the sensor; it useful
to define also the reflectance as the ratio of reflected versus total power energy.
The spectral signature is the reflectance as a function of wavelength (see Figure Spectral Reflectance Curves
of Four Different Targets (page 127)); each material has a unique signature, therefore it can be used for material
classification (NASA, 2013).
Land cover is the material at the ground, such as soil, vegetation, water, asphalt, etc. (Fisher and Unwin, 2005).
Depending on the sensor resolutions, the number and kind of land cover classes that can be identified in the image
can vary significantly.
There are several satellites with different characteristics that acquire multispectral images of earth surface. The
following satellites are particularly useful for land cover monitoring because images are provided for free and can
be downloaded directly from SCP; data have been acquired for the past few decades and the archive is continously
growing with recent images.
Landsat is a set of multispectral satellites developed by the NASA (National Aeronautics and Space Administra-
tion of USA), since the early 1970’s.
Landsat images are very used for environmental research. The resolutions of Landsat 4 and Landsat 5 sensors
are reported in the following table (from http://landsat.usgs.gov/band_designations_landsat_satellites.php); also,
Landsat temporal resolution is 16 days (NASA, 2013).
Landsat 4 and Landsat 5 Bands
The resolutions of Landsat 7 sensor are reported in the following table (from http://landsat.usgs.gov/band_
designations_landsat_satellites.php); also, Landsat temporal resolution is 16 days (NASA, 2013).
Landsat 7 Bands
The resolutions of Landsat 8 sensor are reported in the following table (from http://landsat.usgs.gov/band_
designations_landsat_satellites.php); also, Landsat temporal resolution is 16 days (NASA, 2013).
Landsat 8 Bands
A vast archive of images is freely available from the U.S. Geological Survey . For more information about how to
freely download Landsat images read this .
Images are identified with the paths and rows of the WRS (Worldwide Reference System for Landsat ).
Sentinel-2 is a multispectral satellite developed by the European Space Agency (ESA) in the frame of Copernicus
land monitoring services. Sentinel-2 acquires 13 spectral bands with the spatial resolution of 10m, 20m and 60m
depending on the band, as illustrated in the following table (ESA, 2015).
Sentinel-2 Bands
Sentinel-2 images are freely available from the ESA website https://scihub.copernicus.eu.
Sentinel-3 is a satellite developed by the European Space Agency (ESA) in the frame of Copernicus land mon-
itoring services. It carries several instruments, in particular the Ocean and Land Colour Instrument (OLCI) is a
push-broom imaging spectrometer acquiring 21 bands in the range 0.4-1.02 𝜇m with a swath width of 1,270km
and 300m spatial resolution (ESA, 2013). The revisit time is about 2 days.
Sentinel-3 Bands
The ASTER (Advanced Spaceborne Thermal Emission and Reflection Radiometer) satellite was launched in
1999 by a collaboration between the Japanese Ministry of International Trade and Industry (MITI) and the NASA.
ASTER has 14 bands whose spatial resolution varies with wavelength: 15m in the visible and near-infrared, 30m
in the short wave infrared, and 90m in the thermal infrared (USGS, 2015). ASTER bands are illustrated in the
following table (due to a sensor failure SWIR data acquired since April 1, 2008 is not available ). An additional
band 3B (backwardlooking near-infrared) provides stereo coverage.
ASTER Bands
The MODIS (Moderate Resolution Imaging Spectroradiometer) is an instrument operating on the Terra and Aqua
satellites launched by NASA in 1999 and 2002 respectively. Its temporal resolutions allows for viewing the entire
Earth surface every one to two days, with a swath width of 2,330km. Its sensors measure 36 spectral bands at
three spatial resolutions: 250m, 500m, and 1,000m (see https://lpdaac.usgs.gov/dataset_discovery/modis).
Several products are available, such as surface reflectance and vegetation indices. In this manual we are con-
sidering the surface reflectance bands available at 250m and 500m spatial resolution (Vermote, Roger, & Ray,
2015).
MODIS Bands
The Geostationary Operational Environmental Satellite-R Series (GOES-R) are geostationary satellites de-
veloped for weather monitoring by the National Oceanic and Atmospheric Administration (NOAA) and the NASA
(NOAA, 2020).
GOES constellation is composed of GOES-R satellite (also known as GOES-16 that replaced GOES-13 on
December 2017), and GOES-S satellite (also known as GOES-17, operational since February 2019). Other
satellites (GOES-T and GOES-U) are planned to be launched in the future. For more information please visit
https://www.goes-r.gov/mission/mission.html .
GOES geostationary satellites monitor continuously the same area, a very large portion of Earth surface with three
geographic coverage regions: Full Disk, Continental United States (CONUS), and Mesoscale. In particular, Full
Disk products have hemispheric coverage of 83° local zenith angle, and images are acquired every 5-15 minutes.
GOES-16 monitors from 75.2 degrees west longitude, including America, the Atlantic Ocean, and the west coast
of Africa. GOES-17 monitors from 137.2 degrees west longitude, including the Pacific Ocean.
GOES sensors include several spectral bands; the following table describes the bands downloaded using SCP.
GOES Bands
Synthetic Aperture Radar (SAR) is a technique of active remote sensing that is the sensor platform emits mi-
crowaves in order to acquire images of the ground (Richards and Jia, 2006). In fact, the sensor platform emits
the radiation (at a specific wavelength) and measures the magnitude and the phase of radiation that bounces back
from the ground to the sensor.
Unlike passive sensors, SAR systems can work day and night and can penetrate clouds allowing for the monitoring
of surface also with adverse meteorological conditions; depending on the microwave wavelength, the radiation can
penetrate different types of materials allowing for different applications (NASA, 2020).
The main SAR systems can be divided according to the wavelength as illustrated in the following table (NASA,
2020):
Main SAR Bands
Usually, SAR sensors can emit and measure different polarizations (i.e. orientation of the microwaves of the
electric field), for instance vertical (i.e. polarization oriented in the vertical direction in antenna coordinates) and
horizontal (i.e. polarization oriented in the horizontal direction in antenna coordinates) (ESA, 2020).
SAR systems can acquire in both ascending and descending orbits, however the acquired images are affected by
the different acquisition geometries, which should be considered when mixing ascending and descending images.
Acquisitions are called swaths and usually they are composed of sub-swaths. With particular acquisition modes,
the resolution of pixels along track (the side parallel to the flight direction) can be different than slant-range (the
side perpendicular to the flight direction).
SAR phase information is used to perform interferometry (also InSAR) to measure the distance from the sensor
to the target (NASA, 2020).
For more information, please read the ESA introduction to SAR and the NASA definition of SAR .
Sentinel-1 is a Copernicus mission of satellites that operate at C-band to provide SAR imagery at medium reso-
lution (about 10m).
The Sentinel-1 constellation provides high revisit time (about 5 days), a wide swath (250 km), and acquires
images in different operational modes. The primary operation mode on land is the Interferometric Wide swath
(IW), which is data is acquired in three swaths using the Terrain Observation with Progressive Scanning SAR
(TOPSAR) imaging technique (ESA, 2020b).
The Level-1 products systematically delivered by Copernicus are Single Look Complex (SLC, data comprising
complex imagery with amplitude and phase) and Ground Range Detected (GRD, Level-1 data with multi-looked
intensity only).
Sentinel-1 supports dual polarization, which are horizontal (H) or vertical (V); VV and VH polarimetric channels
are available to classify and analyze land cover such as built-up areas or vegetation.
Sentinel-1 preprocessing
SCP preprocesses Sentinel-1 GRD images performing basically the same methodology applied for the images
included in Google Earth Engine .
The preprocessing comprises the following phases:
• Application of the orbit file;
• Removal of GRD border noise for low intensity and invalid data;
• Removal of thermal noise to reduce discontinuities between sub-swaths;
• Calculation of backscatter intensity using radiometric calibration;
• Orthorectification (terrain correction) using the SRTM 30 meter DEM;
• Conversion of backscatter coefficient to dB.
These processes are performed through the SNAP Graph Processing Tool (GPT) and a graph file .xml . Therefore,
SCP requires ESA SNAP to perform the preprocessing. For the installation of ESA SNAP, please visit this page
http://step.esa.int/main/download/snap-download/ .
The path to the SNAP gpt executable must be entered in settings Processing (page 110). Usually C:/snap/bin/
gpt.exe for Windows, /Applications/snap/bin/gpt for Mac, and /usr/local/snap/bin/gpt
for Linux.
It is also possible to use custom graph files and select to project the raster to the same coordinate reference system
as a band set (default WGS 84).
A semi-automatic classification (also supervised classification) is an image processing technique that allows
for the identification of materials in an image, according to their spectral signatures. There are several kinds of
classification algorithms, but the general purpose is to produce a thematic map of the land cover.
Image processing and GIS spatial analyses require specific software such as the Semi-Automatic Classification
Plugin for QGIS.
Often, a combination is created of three individual monochrome images, in which each is assigned a given color;
this is defined color composite and is useful for photo interpretation (NASA, 2013). Color composites are usually
expressed as:
“R G B = Br Bg Bb”
where:
• R stands for Red;
• G stands for Green;
• B stands for Blue;
• Br is the band number associated to the Red color;
• Bg is the band number associated to the Green color;
• Bb is the band number associated to the Blue color.
The following Figure Color composite of a Landsat 8 image (page 134) shows a color composite “R G B = 4 3 2”
of a Landsat 8 image (for Landsat 7 the same color composite is R G B = 3 2 1; for Sentinel-2 is R G B = 4 3 2) and
a color composite “R G B = 5 4 3” (for Landsat 7 the same color composite is R G B = 4 3 2; for Sentinel-2 is R G
B = 8 4 3). The composite “R G B = 5 4 3” is useful for the interpretation of the image because vegetation pixels
appear red (healthy vegetation reflects a large part of the incident light in the near-infrared wavelength, resulting
in higher reflectance values for band 5, thus higher values for the associated color red).
Usually, supervised classifications require the user to select one or more Regions of Interest (ROIs, also Training
Areas) for each land cover class identified in the image. ROIs are polygons drawn over homogeneous areas of the
image that overlay pixels belonging to the same land cover class.
The Region Growing Algorithm allows to select pixels similar to a seed one, considering the spectral similarity
(i.e. spectral distance) of adjacent pixels. In SCP the Region Growing Algorithm is available for the training area
creation. The parameter distance is related to the similarity of pixel values (the lower the value, the more similar
are selected pixels) to the seed one (i.e. selected clicking on a pixel). An additional parameter is the maximum
width, which is the side length of a square, centred at the seed pixel, which inscribes the training area (if all the
pixels had the same value, the training area would be this square). The minimum size is used a constraint (for
every single band), selecting at least the pixels that are more similar to the seed one until the number of selected
pixels equals the minimum size.
In figure Region growing example (page 135) the central pixel is used as seed (image a) for the region growing
of one band (image b) with the parameter spectral distance = 0.1; similar pixels are selected to create the training
area (image c and image d).
Land cover classes are identified with an arbitrary ID code (i.e. Identifier). SCP allows for the definition of
Macroclass ID (i.e. MC ID) and Class ID (i.e. C ID), which are the identification codes of land cover classes.
A Macroclass is a group of ROIs having different Class ID, which is useful when one needs to classify materials
that have different spectral signatures in the same land cover class. For instance, one can identify grass (e.g. ID
class = 1 and Macroclass ID = 1 ) and trees (e.g. ID class = 2 and Macroclass ID = 1 ) as
vegetation class (e.g. Macroclass ID = 1 ). Multiple Class IDs can be assigned to the same Macroclass ID,
but the same Class ID cannot be assigned to multiple Macroclass IDs, as shown in the following table.
Example of Macroclasses
Therefore, Classes are subsets of a Macroclass as illustrated in Figure Macroclass example (page 136).
If the use of Macroclass is not required for the study purpose, then the same Macroclass ID can be defined for all
the ROIs (e.g. Macroclass ID = 1) and Macroclass values are ignored in the classification process.
The spectral signatures (spectral characteristics) of reference land cover classes are calculated considering the
values of pixels under each ROI having the same Class ID (or Macroclass ID). Therefore, the classification algo-
rithm classifies the whole image by comparing the spectral characteristics of each pixel to the spectral character-
istics of reference land cover classes. SCP implements the following classification algorithms.
Minimum Distance
Minimum Distance algorithm calculates the Euclidean distance 𝑑(𝑥, 𝑦) between spectral signatures of image pixels
and training spectral signatures, according to the following equation:
⎯
⎸ 𝑛
⎸∑︁
𝑑(𝑥, 𝑦) = ⎷ (𝑥𝑖 − 𝑦𝑖 )2
𝑖=1
where:
• 𝑥 = spectral signature vector of an image pixel;
• 𝑦 = spectral signature vector of a training area;
where:
• 𝐶𝑘 = land cover class 𝑘;
• 𝑦𝑘 = spectral signature of class 𝑘;
• 𝑦𝑗 = spectral signature of class 𝑗.
It is possible to define a threshold 𝑇𝑖 in order to exclude pixels below this value from the classification:
𝑥 ∈ 𝐶𝑘 ⇐⇒ 𝑑(𝑥, 𝑦𝑘 ) < 𝑑(𝑥, 𝑦𝑗 )∀𝑘 ̸= 𝑗
𝑎𝑛𝑑
𝑑(𝑥, 𝑦𝑘 ) < 𝑇𝑖
Maximum Likelihood
Maximum Likelihood algorithm calculates the probability distributions for the classes, related to Bayes’ theorem,
estimating if a pixel belongs to a land cover class. In particular, the probability distributions for the classes are
assumed the of form of multivariate normal models (Richards & Jia, 2006). In order to use this algorithm, a
sufficient number of pixels is required for each training area allowing for the calculation of the covariance matrix.
The discriminant function, described by Richards and Jia (2006), is calculated for every pixel as:
1 1
𝑔𝑘 (𝑥) = ln 𝑝(𝐶𝑘 ) − ln |Σ𝑘 | − (𝑥 − 𝑦𝑘 )𝑡 Σ−1
𝑘 (𝑥 − 𝑦𝑘 )
2 2
where:
• 𝐶𝑘 = land cover class 𝑘;
• 𝑥 = spectral signature vector of a image pixel;
• 𝑝(𝐶𝑘 ) = probability that the correct class is 𝐶𝑘 ;
• |Σ𝑘 | = determinant of the covariance matrix of the data in class 𝐶𝑘 ;
• Σ−1
𝑘 = inverse of the covariance matrix;
In addition, it is possible to define a threshold to the discriminant function in order to exclude pixels below this
value from the classification. Considering a threshold 𝑇𝑖 the classification condition becomes:
𝑥 ∈ 𝐶𝑘 ⇐⇒ 𝑔𝑘 (𝑥) > 𝑔𝑗 (𝑥)∀𝑘 ̸= 𝑗
𝑎𝑛𝑑
𝑔𝑘 (𝑥) > 𝑇𝑖
Maximum likelihood is one of the most common supervised classifications, however the classification process can
be slower than Minimum Distance (page 135).
The Spectral Angle Mapping calculates the spectral angle between spectral signatures of image pixels and training
spectral signatures. The spectral angle 𝜃 is defined as (Kruse et al., 1993):
(︃ ∑︀𝑛 )︃
−1 𝑖=1 𝑥𝑖 𝑦𝑖
𝜃(𝑥, 𝑦) = cos ∑︀𝑛 1 ∑︀𝑛 1
( 𝑖=1 𝑥2𝑖 ) 2 * ( 𝑖=1 𝑦𝑖2 ) 2
Where:
where:
• 𝐶𝑘 = land cover class 𝑘;
• 𝑦𝑘 = spectral signature of class 𝑘;
• 𝑦𝑗 = spectral signature of class 𝑗.
In order to exclude pixels below this value from the classification it is possible to define a threshold 𝑇𝑖 :
Parallelepiped Classification
Parallelepiped classification is an algorithm that considers a range of values for each band, forming a multidi-
mensional parallelepiped that defines a land cover class. A pixel is classified if the values thereof are inside a
parallelepiped. One of the major drawbacks is that pixels whose signatures lie in the overlapping area of two or
more parallelepipeds cannot be classified (Richards and Jia, 2006).
Land Cover Signature Classification is available in SCP (see Land Cover Signature Classification (page 72)). This
classification allows for the definition of spectral thresholds for each training input signature (a minimum value
and a maximum value for each band). The thresholds of each training input signature define a spectral region
belonging to a certain land cover class.
Spectral signatures of image pixels are compared to the training spectral signatures; a pixel belongs to class X
if pixel spectral signature is completely contained in the spectral region defined by class X. In case of pixels
falling inside overlapping regions or outside any spectral region, it is possible to use additional classification algo-
rithms (i.e. Minimum Distance (page 135), Maximum Likelihood (page 136), Spectral Angle Mapping (page 137))
considering the spectral characteristics of the original input signature.
In the following image, a scheme illustrates the Land Cover Signature Classification for a simple case of two
spectral bands 𝑥 and 𝑦. User defined spectral regions define three classes (𝑔𝑎 , 𝑔𝑏 , and 𝑔𝑐 ). Point 𝑝1 belongs to
class 𝑔𝑎 and point 𝑝2 belongs to class 𝑔𝑏 . However, point 𝑝3 is inside the spectral regions of both classes 𝑔𝑏
and 𝑔𝑐 (overlapping regions); in this case, point 𝑝3 will be unclassified or classified according to an additional
classification algorithm. Point 𝑝4 is outside any spectral region, therefore it will be unclassified or classified
according to an additional classification algorithm. Given that point 𝑝4 belongs to class 𝑔𝑐 , the spectral region
thereof could be extended to include point 𝑝4 .
This is similar to Parallelepiped Classification (page 138), with the exception that spectral regions are defined by
user, and can be assigned independently for the upper and lower bounds. One can imagine spectral regions as the
set of all the spectral signatures of pixels belonging to one class.
In figure Plot of spectral ranges (page 140) the spectral ranges of three classes (𝑔𝑎 , 𝑔𝑏 , and 𝑔𝑐 ) are displayed; the
colored lines inside the ranges (i.e. semi-transparent area) represent the spectral signatures of pixels that defined
the upper and lower bounds of the respective ranges. Pixel 𝑝1 (dotted line) belongs to class 𝑔𝑏 because the spectral
signature thereof is completely inside the range of class 𝑔𝑏 (in the upper limit); pixel 𝑝2 (dashed line) is unclassified
because the spectral signature does not fall completely inside any range; pixel 𝑝3 (dotted line) belongs to class 𝑔𝑎 .
It is worth noticing that these spectral thresholds can be applied to any spectral signature, regardless of spectral
characteristics thereof; this function can be very useful for separating similar spectral signatures that differ only
in one band, defining thresholds that include or exclude specific signatures. In fact, classes are correctly separated
if the spectral ranges thereof are not overlapping at least in one band. Of course, even if spectral regions are
overlapping, chances are that no pixel will fall inside the overlapping region and be misclassified; which is the
upper (or lower) bound of a range do not imply the existence, in the image, of any spectral signature having the
maximum (or minimum) range values for all the bands (for instance pixel 𝑝1 of figure Plot of spectral ranges
(page 140) could not exist).
One of the main benefit of the Land Cover Signature Classification is that it is possible to select pixels and and
include the signature thereof in a spectral range; therefore, the classification should be the direct representation of
the class expected for every spectral signature. This is very suitable for the classification of a single land cover
class (defined by specific spectral thresholds), and leave unclassified the rest of the image that is of no interest for
the purpose of the classification.
Algorithm raster
An algorithm raster represents the “distance” (according to the definition of the classification algorithm) of an
image pixel to a specific spectral signature.
In general, an algorithm raster is produced for every spectral signature used as training input. The value of every
pixel is the result of the algorithm calculation for a specific spectral signature. Therefore, a pixel belongs to class
X if the value of the algorithm raster corresponding to class X is the lowest in case of Minimum Distance
(page 135) or Spectral Angle Mapping (page 137) (or highest in case of Maximum Likelihood (page 136)).
Given a classification, a combination of algorithm rasters can be produced, in order to create a raster with the
lowest “distances” (i.e. pixels have the value of the algorithm raster corresponding to the class they belong in the
classification). Therefore, this raster can be useful to identify pixels that require the collection of more similar
spectral signatures (see Classification preview (page 30)).
Machine Learning is a broad set of classification techniques that aim to build mathematical models based on
training data.
In general, Machine Learning algorithms split the data in (ESA, 2019):
• Training Dataset: the sample data used to fit the model;
• Validation Dataset: the sample data used to tune the model parameters to fit on the training dataset;
• Test Dataset: the sample data used to provide an evaluation of the final model;
Usually, the training and model evaluation are performed iteratively.
Random Forest
Random Forest is a particular machine learning technique, based on the iterative and random creation of decision
trees (i.e. a set of rules and conditions that define a class).
First, the input features should be defined, which can be spectral bands or ancillary rasters. Training Areas
(page 133) should be created to define the classes used for training the model.
Random Forest calculates several random decision trees, based on the following parameters:
• number of training samples: is the number of training data (pixels) randomly used to train the model; it
should be set lower than total training input pixels;
• number of trees: is the number of decision trees; the more the number of trees, the more is the model
accuracy, but it also increases the calculation time.
For instance, a decision tree could be defined as:
• class 1 = band 1 > 0.1 –> band 2 < 0.3 –> band 3 > 0.4
• class 2 = band 1 > 0.4 –> band 2 > 0.6 –> band 3 < 0.1
• class 3 = band 1 < 0.7 –> band 2 > 0.1 –> band 3 < 0.5
Random Forest creates several decision trees randomly. Usually, the Gini coefficient is calculated to split the trees.
Therefore, a model based on the decision trees is created and used to classify all the pixels.
A pixel is classified according to the majority vote of decision trees, for example a pixel is classified as class 1 if
most decision trees evaluated it as class 1. Also, a confidence layer is produced, which measures the uncertainty
of the model based on training data.
Random Forest can be used to evaluate the importance of input features, according to the contribution thereof to
the model.
It is useful to evaluate the spectral distance (or separability) between training signatures or pixels, in order to
assess if different classes that are too similar could cause classification errors. The SCP implements the following
algorithms for assessing similarity of spectral signatures.
Jeffries-Matusita Distance
Jeffries-Matusita Distance calculates the separability of a pair of probability distributions. This can be particularly
meaningful for evaluating the results of Maximum Likelihood (page 136) classifications.
The Jeffries-Matusita Distance 𝐽𝑥𝑦 is calculated as (Richards and Jia, 2006):
𝐽𝑥𝑦 = 2 1 − 𝑒−𝐵
(︀ )︀
where:
)︂−1 (︃ Σ𝑥 +Σ𝑦
)︃
| |
(︂
1 Σ𝑥 + Σ 𝑦 1
𝐵 = (𝑥 − 𝑦)𝑡 (𝑥 − 𝑦) + ln 1
2
1
8 2 2 |Σ𝑥 | 2 |Σ𝑦 | 2
where:
• 𝑥 = first spectral signature vector;
• 𝑦 = second spectral signature vector;
• Σ𝑥 = covariance matrix of sample 𝑥;
• Σ𝑦 = covariance matrix of sample 𝑦;
The Jeffries-Matusita Distance is asymptotic to 2 when signatures are completely different, and tends to 0 when
signatures are identical.
Spectral Angle
The Spectral Angle is the most appropriate for assessing the Spectral Angle Mapping (page 137) algorithm. The
spectral angle 𝜃 is defined as (Kruse et al., 1993):
(︃ ∑︀𝑛 )︃
−1 𝑖=1 𝑥𝑖 𝑦𝑖
𝜃(𝑥, 𝑦) = cos ∑︀𝑛 1 ∑︀𝑛 1
( 𝑖=1 𝑥2𝑖 ) 2 * ( 𝑖=1 𝑦𝑖2 ) 2
Where:
• 𝑥 = spectral signature vector of an image pixel;
• 𝑦 = spectral signature vector of a training area;
• 𝑛 = number of image bands.
Spectral angle goes from 0 when signatures are identical to 90 when signatures are completely different.
Euclidean Distance
The Euclidean Distance is particularly useful for the evaluating the result of Minimum Distance (page 135) clas-
sifications. In fact, the distance is defined as:
⎯
⎸ 𝑛
⎸∑︁
𝑑(𝑥, 𝑦) = ⎷ (𝑥𝑖 − 𝑦𝑖 )2
𝑖=1
where:
• 𝑥 = first spectral signature vector;
• 𝑦 = second spectral signature vector;
• 𝑛 = number of image bands.
The Euclidean Distance is 0 when signatures are identical and tends to increase according to the spectral distance
of signatures.
Bray-Curtis Similarity
The Bray-Curtis Similarity is a statistic used for assessing the relationship between two samples (read this). It is
useful in general for assessing the similarity of spectral signatures, and Bray-Curtis Similarity 𝑆(𝑥, 𝑦) is calculated
as:
(︂ ∑︀𝑛 )︂
𝑖=1 |(𝑥𝑖 − 𝑦𝑖 )|
𝑆(𝑥, 𝑦) = 100 − ∑︀𝑛 ∑︀𝑛 * 100
𝑖=1 𝑥𝑖 + 𝑖=1 𝑦𝑖
where:
The result of the classification process is a raster (see an example of Landsat classification in Figure Landsat
classification (page 143)), where pixel values correspond to class IDs and each color represent a land cover class.
A certain amount of errors can occur in the land cover classification (i.e. pixels assigned to a wrong land cover
class), due to spectral similarity of classes, or wrong class definition during the ROI collection.
After the classification process, it is useful to assess the accuracy of land cover classification, in order to identify
and measure map errors. Usually, accuracy assessment is performed with the calculation of an error matrix,
which is a table that compares map information with reference data (i.e. ground truth data) for a number of
sample areas (Congalton and Green, 2009).
The following table is a scheme of error matrix, where k is the number of classes identified in the land cover
classification, and n is the total number of collected sample units. The items in the major diagonal (aii) are the
number of samples correctly identified, while the other items are classification error.
Scheme of Error Matrix
Therefore, it is possible to calculate the overall accuracy as the ratio between the number of samples that are
correctly classified (the sum of the major diagonal), and the total number of sample units n (Congalton and Green,
2009).
The overall accuracy (also expressed in percentage) is defined as:
𝑘
∑︁
𝑂= 𝑎𝑖𝑖 /𝑛
𝑖=1
The user’s accuracy for each class is defined as the ratio (also expressed in percentage) between correct samples
and the row total:
𝑈𝑖 = 𝑎𝑖𝑖 /𝑎𝑖+
The commission error 𝐶𝐸𝑖 = 1 − 𝑈𝑖 corresponds to pixels classified as class 𝑖 that actually belong to a different
class.
The producer’s accuracy for each class is calculated as the ratio (also expressed in percentage) between correct
samples and the column total:
𝑃𝑖 = 𝑎𝑖𝑖 /𝑎+𝑖
The omission error 𝑂𝐸𝑖 = 1 − 𝑃𝑖 corresponds to pixels actually belonging to class 𝑖 that were classified erro-
neously as a different class.
It is recommended to calculate the area based error matrix (Olofsson et al., 2014) where each element represents
the estimated area proportion of each class. This allows for estimating the unbiased user’s accuracy and producer’s
accuracy, the unbiased area of classes according to reference data, and the standard error of area estimates.
For further information, the following documentation is freely available: Landsat Data Users Handbook.
Remote sensing images can be processed in various ways in order to obtain classification, indices, or other derived
information that can be useful for land cover characterization.
Principal Component Analysis (PCA) is a method for reducing the dimensions of measured variables (bands) to
the principal components (JARS, 1993).
Th principal component transformation provides a new set of bands (principal components) having the following
characteristic: principal components are uncorrelated; each component has variance less than the previous com-
ponent. Therefore, this is an efficient method for extracting information and data compression (Ready and Wintz,
1973).
Given an image with N spectral bands, the principal components are obtained by matrix calculation (Ready and
Wintz, 1973; Richards and Jia, 2006):
𝑌 = 𝐷𝑡 𝑋
where:
• 𝑌 = vector of principal components
• 𝐷 = matrix of eigenvectors of the covariance matrix 𝐶𝑥 in X space
• 𝑡 denotes vector transpose
And 𝑋 is calculated as:
𝑋 =𝑃 −𝑀
Usually the first two components contain more than the 90% of the variance. For example, the first principal
components can be displayed in a Color Composite (page 133) for highlighting Land Cover (page 126) classes, or
used as input for Supervised Classification (page 133).
4.5.2 Pan-sharpening
Pan-sharpening is the combination of the spectral information of multispectral bands (MS), which have lower
spatial resolution (for Landsat bands, spatial resolution is 30m), with the spatial resolution of a panchromatic
band (PAN), which for Landsat 7 and 8 it is 15m. The result is a multispectral image with the spatial resolution
of the panchromatic band (e.g. 15m). In SCP, a Brovey Transform is applied, where the pan-sharpened values of
each multispectral band are calculated as (Johnson, Tateishi and Hoan, 2012):
𝑀 𝑆𝑝𝑎𝑛 = 𝑀 𝑆 * 𝑃 𝐴𝑁/𝐼
Fig. 12: Example of pan-sharpening of a Landsat 8 image. Left, original multispectral bands (30m); right, pan-
sharpened bands (15m)
Data available from the U.S. Geological Survey
Spectral indices are operations between spectral bands that are useful for extracting information such as vegetation
cover (JARS, 1993). One of the most popular spectral indices is the Normalized Difference Vegetation Index
(NDVI), defined as (JARS, 1993):
𝑁 𝐷𝑉 𝐼 = (𝑁 𝐼𝑅 − 𝑅𝑒𝑑)/(𝑁 𝐼𝑅 + 𝑅𝑒𝑑)
NDVI values range from -1 to 1. Dense and healthy vegetation show higher values, while non-vegetated areas
show low NDVI values.
Another index is the Enhanced Vegetation Index (EVI) which attempts to account for atmospheric effects such
as path radiance calculating the difference between the blue and the red bands (Didan,et al., 2015). EVI is defined
as:
where: 𝐺 is a scaling factor, 𝐶1 and 𝐶2 are coefficients for the atmospheric effects, and 𝐿 is a factor for accounting
the differential NIR and Red radiant transfer through the canopy. Typical coefficient values are: 𝐺 = 2.5, 𝐿 = 1,
𝐶1 = 6, 𝐶2 = 7.5 (Didan,et al., 2015).
4.5.4 Clustering
Clustering is the grouping of pixels based on spectral similarity (e.g. Euclidean Distance (page 142) or Spectral
Angle (page 142)) calculated for a multispectral image (Richards and Jia, 2006).
Clustering can be used for unsupervised classification or for the automatic selection of spectral signatures. It is
worth noticing that, while Supervised Classification (page 133) produces a classification whit the classes identified
during the training process, the classes produced by clustering (i.e. clusters) have no definition and consequently
the user must assign a land cover label to each class.
The main advantage of clustering resides in automation. Of course, clusters do not necessarily represent a partic-
ular land cover type and additional processing could be required for producing an accurate classification.
There are several types of clustering, mainly based on iterative methods; the following are the algorithms provided
in SCP.
K-means
The K-means method is based on the calculation of the average spectral signature of clusters (Wikipedia, 2017;
JARS, 1993).
At first, the user defines the number of clusters expected in the image, which correspond to as many spectral
signatures (i.e. seeds). Starting spectral signatures can be selected in various ways (e.g. randomly, provided by
the user, calculated automatically from image values).
During the first iteration clusters are produced calculating the pixel spectral distance with initial spectral sig-
natures. The algorithms Euclidean Distance (page 142) or Spectral Angle (page 142) can be used for distance
calculation. Pixels are assigned according to the most similar spectral signature, therefore producing clusters.
Then, the average spectral signature is calculated for each cluster of pixels, resulting in the spectral signatures that
will be used in the following iteration.
This process continues iteratively producing clusters and mean spectral signatures, until one of the following
condition is verified:
• the spectral distance between the spectral signatures produced in this iteration with the corresponding ones
produced in the previous iteration is lower than a certain threshold;
• the maximum number of iterations is reached.
After the last iteration, a raster of clusters is produced using the spectral signatures derived from the last iteration.
ISODATA
The ISODATA (Iterative Self-Organizing Data Analysis Technique) method is similar to K-means but with the
additional steps of merging clusters having similar spectral signatures and splitting clusters having too high vari-
ability (i.e. standard deviation) of spectral signatures (Ball & Hall, 1965). Following, the SCP implementation of
ISODATA is described.
At first, the user defines the number of clusters expected in the image, which correspond to as many spectral
signatures (i.e. seeds). Starting spectral signatures can be selected in various ways (e.g. randomly, provided by
the user, calculated automatically from image values). Initial parameters provided by user are:
• 𝐶 = number of desired clusters
• 𝑁𝑚𝑖𝑛 = minimum number of pixels for a cluster
• 𝜎𝑡 = maximum standard deviation threshold for splitting
• 𝐷𝑡 = distance threshold for merging
During the first iteration clusters are produced calculating the Euclidean Distance (page 142) of pixels with initial
spectral signatures. Pixels are assigned according to the most similar spectral signature, therefore producing
clusters.
Therefore, the following parameters are calculated:
• 𝑁𝑖 = number of pixels of cluster 𝑖
• 𝑆𝑖 = average spectral signature of cluster 𝑖
• 𝐴𝑉 𝐸𝑅𝐴𝐺𝐸𝐷𝐼𝑆𝑇𝑖 = average distance of cluster 𝑖 with the seed spectral signature
• 𝐴𝑉 𝐸𝑅𝐴𝐺𝐸𝐷𝐼𝑆𝑇 𝐴𝑁 𝐶𝐸 = overall average distance of all clusters
• 𝜎𝑖𝑗 = standard deviation of cluster 𝑖 in band 𝑗
• 𝜎𝑚𝑎𝑥𝑖 = maximum standard deviation of cluster 𝑖 (i.e. 𝑚𝑎𝑥(𝜎𝑖𝑗 ))
• 𝑘𝑖 = band where 𝜎𝑚𝑎𝑥𝑖 occurred
• 𝑆𝑘𝑖 = value of 𝑆𝑖 at band 𝑘𝑖
• 𝑃 = number of clusters
Then, for each cluster 𝑖, if 𝑁𝑖 < 𝑁𝑚𝑖𝑛 , then the cluster 𝑖 is discarded.
If 𝑃 <= 𝐶 then try to split clusters. For each cluster 𝑖:
• If 𝜎𝑚𝑎𝑥𝑖 > 𝜎𝑡 :
– If ((𝐴𝑉 𝐸𝑅𝐴𝐺𝐸𝐷𝐼𝑆𝑇𝑖 > 𝐴𝑉 𝐸𝑅𝐴𝐺𝐸𝐷𝐼𝑆𝑇 𝐴𝑁 𝐶𝐸) AND (𝑁𝑖 > (2 * 𝑁𝑚𝑖𝑛 + 2) )) OR (𝐶 > 2 *
𝑃 ):
This chapter provides information about the conversion to reflectance implemented in SCP.
Radiance is the “flux of energy (primarily irradiant or incident energy) per solid angle leaving a unit surface area
in a given direction”, “Radiance is what is measured at the sensor and is somewhat dependent on reflectance”
(NASA, 2011, p. 47).
Images such as Landsat or Sentinel-2 are composed of several bands and a metadata file which contains informa-
tion required for the conversion to reflectance.
Landsat images are provided in radiance, scaled prior to output. For Landsat images Spectral Radiance at the
sensor’s aperture (𝐿𝜆 , measured in [watts/(meter squared * ster * 𝜇𝑚)]) is given by (https://www.usgs.gov/
core-science-systems/nli/landsat/using-usgs-landsat-level-1-data-product):
𝐿𝜆 = 𝑀𝐿 * 𝑄𝑐𝑎𝑙 + 𝐴𝐿
where:
• 𝑀𝐿 = Band-specific multiplicative rescaling factor from Landsat metadata (RADI-
ANCE_MULT_BAND_x, where x is the band number)
• 𝐴𝐿 = Band-specific additive rescaling factor from Landsat metadata (RADIANCE_ADD_BAND_x, where
x is the band number)
• 𝑄𝑐𝑎𝑙 = Quantized and calibrated standard product pixel values (DN)
Sentinel-2 images (Level-1C) are already provided in Top Of Atmosphere (TOA) Reflectance (page 148), scaled
prior to output (ESA, 2015).
Images in radiance can be converted to Top Of Atmosphere (TOA) Reflectance (combined surface and atmospheric
reflectance) in order to reduce the in between-scene variability through a normalization for solar irradiance. This
TOA reflectance (𝜌𝑝 ), which is the unitless ratio of reflected versus total power energy (NASA, 2011), is calculated
by:
𝜌𝑝 = (𝜋 * 𝐿𝜆 * 𝑑2 )/(𝐸𝑆𝑈 𝑁𝜆 * 𝑐𝑜𝑠𝜃𝑠 )
where:
• 𝐿𝜆 = Spectral radiance at the sensor’s aperture (at-satellite radiance)
• 𝑑 = Earth-Sun distance in astronomical units (provided with Landsat 8 metadata file, and an excel file is
available from http://landsathandbook.gsfc.nasa.gov/excel_docs/d.xls)
• 𝐸𝑆𝑈 𝑁𝜆 = Mean solar exo-atmospheric irradiances
• 𝜃𝑠 = Solar zenith angle in degrees, which is equal to 𝜃𝑠 = 90° - 𝜃𝑒 where 𝜃𝑒 is the Sun elevation
It is worth pointing out that Landsat 8 images are provided with band-specific rescaling factors that allow for the
direct conversion from DN to TOA reflectance.
Sentinel-2 images are already provided in scaled TOA reflectance, which can be converted to TOA reflectance
with a simple calculation using the Quantification Value provided in the metadata (see https://sentinel.esa.int/
documents/247904/349490/S2_MSI_Product_Specification.pdf).
Sentinel-3 images are already provided in scaled TOA radiance. Conversion to reflectance is performed applying
the coefficients scale_factor and add_offset provided in the metadata of each band. The ancillary raster
tie_geometries.nc provides the value of sun zenith angle and the ancillary raster instrument_data
provides information about the solar flux for each band, which are used for the conversion to reflectance with
the correction for sun angle. In addition, the georeferencing of the bands is performed using the ancillary raster
geo_coordinates.nc which provides coordinates of every pixel.
The effects of the atmosphere (i.e. a disturbance on the reflectance that varies with the wavelength) should be
considered in order to measure the reflectance at the ground.
As described by Moran et al. (1992), the land surface reflectance (𝜌) is:
where:
• 𝐿𝑝 is the path radiance
• 𝑇𝑣 is the atmospheric transmittance in the viewing direction
• 𝑇𝑧 is the atmospheric transmittance in the illumination direction
• 𝐸𝑑𝑜𝑤𝑛 is the downwelling diffuse irradiance
Therefore, we need several atmospheric measurements in order to calculate 𝜌 (physically-based corrections).
Alternatively, it is possible to use image-based techniques for the calculation of these parameters, without in-situ
measurements during image acquisition. It is worth mentioning that Landsat Surface Reflectance High Level
Data Products for Landsat 8 are available (for more information read http://landsat.usgs.gov/CDR_LSR.php).
The Dark Object Subtraction (DOS) is a family of image-based atmospheric corrections. Chavez (1996) explains
that “the basic assumption is that within the image some pixels are in complete shadow and their radiances received
at the satellite are due to atmospheric scattering (path radiance). This assumption is combined with the fact that
very few targets on the Earth’s surface are absolute black, so an assumed one-percent minimum reflectance is
better than zero percent”. It is worth pointing out that the accuracy of image-based techniques is generally lower
than physically-based corrections, but they are very useful when no atmospheric measurements are available as
they can improve the estimation of land surface reflectance. The path radiance is given by (Sobrino et al., 2004):
𝐿𝑝 = 𝐿𝑚𝑖𝑛 − 𝐿𝐷𝑂1%
where:
• 𝐿𝑚𝑖𝑛 = “radiance that corresponds to a digital count value for which the sum of all the pixels with digital
counts lower or equal to this value is equal to the 0.01% of all the pixels from the image considered”
(Sobrino et al., 2004, p. 437), therefore the radiance obtained with that digital count value (𝐷𝑁𝑚𝑖𝑛 )
• 𝐿𝐷𝑂1% = radiance of Dark Object, assumed to have a reflectance value of 0.01
In particular for Landsat images:
𝐿𝑚𝑖𝑛 = 𝑀𝐿 * 𝐷𝑁𝑚𝑖𝑛 + 𝐴𝐿
There are several DOS techniques (e.g. DOS1, DOS2, DOS3, DOS4), based on different assumption about 𝑇𝑣 ,
𝑇𝑧 , and 𝐸𝑑𝑜𝑤𝑛 . The simplest technique is the DOS1, where the following assumptions are made (Moran et al.,
1992):
• 𝑇𝑣 = 1
• 𝑇𝑧 = 1
• 𝐸𝑑𝑜𝑤𝑛 = 0
Therefore the path radiance is:
ESUN [W /(m2 * 𝜇𝑚)] values for Landsat sensors are provided in the following table.
ESUN values for Landsat bands
Band Sentinel-2
1 1913.57
2 1941.63
3 1822.61
4 1512.79
5 1425.56
6 1288.32
7 1163.19
8 1036.39
8A 955.19
9 813.04
10 367.15
11 245.59
12 85.25
ESUN [W /(m2 * 𝜇𝑚)] values for ASTER sensor are illustrated in the following table (from Finn et al., 2012).
ESUN values for ASTER bands
Band ASTER
1 1848
2 1549
3 1114
4 225.4
5 86.63
6 81.85
7 74.85
8 66.49
9 59.85
An example of comparison of to TOA reflectance, DOS1 corrected reflectance and the Landsat Surface Reflectance
High Level Data Products (ground truth) is provided in Figure Spectral signatures of a built-up pixel (page 151).
This chapter provides the basic information about the conversion to At-Satellite Brightness Temperature imple-
mented in SCP and the estimation of Land Surface Temperature.
For thermal bands, the conversion of DN to At-Satellite Brightness Temperature is given by (from https://www.
usgs.gov/core-science-systems/nli/landsat/using-usgs-landsat-level-1-data-product ):
𝑇𝐵 = 𝐾2 /𝑙𝑛[(𝐾1 /𝐿𝜆 ) + 1]
where:
• 𝐾1 = Band-specific thermal conversion constant (in watts/meter squared * ster * 𝜇𝑚)
• 𝐾2 = Band-specific thermal conversion constant (in kelvin)
and 𝐿𝜆 is the Spectral Radiance at the sensor’s aperture, measured in watts/(meter squared * ster * 𝜇𝑚).
The 𝐾1 and 𝐾2 constants for Landsat sensors are provided in the following table.
Thermal Conversion Constants for Landsat
𝐾1 = 𝑐1 /𝜆5
𝐾2 = 𝑐2 /𝜆
where (Mohr, Newell, & Taylor, 2015):
• 𝑐1 = first radiation constant = 1.191 * 10−16 𝑊 𝑚2 𝑠𝑟−1
• 𝑐2 = second radiation constant = 1.4388 * 10−2 𝑚𝐾
Therefore, for ASTER bands 𝐾1 and 𝐾2 are provided in the following table.
Thermal Conversion Constants for ASTER
Several studies have described the estimation of Land Surface Temperature. Land Surface Temperature can be
calculated from At-Satellite Brightness Temperature 𝑇𝐵 as (Weng et al., 2004):
where:
• 𝜆 = wavelength of emitted radiance
• 𝑐2 = ℎ * 𝑐/𝑠 = 1.4388 * 10−2 m K
• ℎ = Planck’s constant = 6.626 * 10−34 J s
• 𝑠 = Boltzmann constant = 1.38 * 10−23 J/K
• 𝑐 = velocity of light = 2.998 * 108 m/s
The values of 𝜆 for the thermal bands of Landsat and ASTER satellites can be calculated from the tables in Landsat
Satellites (page 127) and ASTER Satellite (page 129).
Several studies used NDVI for the estimation of land surface emissivity (Sobrino et al., 2004); other studies used
a land cover classification for the definition of the land surface emissivity of each class (Weng et al. 2004). For
instance, the emissivity (𝑒) values of various land cover types are provided in the following table (from Mallick et
al., 2012).
Emissivity values
4.8 References
• Ball, G. H. & Hall, D. J., 1965. ISODATA. A novel method of data analysis and pattern classification.
Menlo Park: Stanford Research Institute.
• Chander, G. & Markham, B., 2003. Revised Landsat-5 TM radiometric calibration procedures and postcal-
ibration dynamic ranges Geoscience and Remote Sensing, IEEE Transactions on, 41, 2674 - 2677
• Chavez, P. S., 1996. Image-Based Atmospheric Corrections - Revisited and Improved Photogrammetric
Engineering and Remote Sensing, [Falls Church, Va.] American Society of Photogrammetry, 62, 1025-
1036
• Congalton, R. and Green, K., 2009. Assessing the Accuracy of Remotely Sensed Data: Principles and
Practices. Boca Raton, FL: CRC Press
• Didan, K.; Barreto Munoz, A.; Solano, R. & Huete, A., 2015. MODIS Vegetation Index User’s Guide.
Collection 6, NASA
• ESA, 2020. Sentinel-1 SAR Definitions. Available at https://sentinel.esa.int/web/sentinel/user-guides/
sentinel-1-sar/definitions
• ESA, 2020b. Sentinel-1 SAR Definitions. Available at https://sentinel.esa.int/web/sentinel/user-guides/
sentinel-1-sar/overview
• ESA, 2019. A machine learning glossary. Available at https://blogs.esa.int/philab/2019/03/29/
a-machine-learning-glossary/
• ESA, 2015. Sentinel-2 User Handbook. Available at https://sentinels.copernicus.eu/documents/247904/
685211/Sentinel-2_User_Handbook
• ESA, 2013. Sentinel-3 User Handbook. Available at https://sentinels.copernicus.eu/documents/247904/
685236/Sentinel-3_User_Handbook
• Finn, M.P., Reed, M.D, and Yamamoto, K.H., 2012. A Straight Forward Guide for Pro-
cessing Radiance and Reflectance for EO-1 ALI, Landsat 5 TM, Landsat 7 ETM+, and
ASTER. Unpublished Report from USGS/Center of Excellence for Geospatial Information Science,
8 p, http://cegis.usgs.gov/soil_moisture/pdf/A%20Straight%20Forward%20guide%20for%20Processing%
20Radiance%20and%20Reflectance_V_24Jul12.pdf
• Fisher, P. F. and Unwin, D. J., eds., 2005. Representing GIS. Chichester, England: John Wiley & Sons
• JARS, 1993. Remote Sensing Note. Japan Association on Remote Sensing. Available at http://www.
jars1974.net/pdf/rsnote_e.html
• Jimenez-Munoz, J. C. & Sobrino, J. A., 2010. A Single-Channel Algorithm for Land-Surface Temperature
Retrieval From ASTER Data IEEE Geoscience and Remote Sensing Letters, 7, 176-179
• Johnson, B. A., Tateishi, R. and Hoan, N. T., 2012. Satellite Image Pansharpening Using a Hybrid Approach
for Object-Based Image Analysis ISPRS International Journal of Geo-Information, 1, 228. Available at
http://www.mdpi.com/2220-9964/1/3/228)
• Kruse, F. A., et al., 1993. The Spectral Image Processing System (SIPS) - Interactive Visualization and
Analysis of Imaging spectrometer. Data Remote Sensing of Environment
• Mallick, J.; Singh, C. K.; Shashtri, S.; Rahman, A. & Mukherjee, S., 2012. Land surface emissivity re-
trieval based on moisture index from LANDSAT TM satellite data over heterogeneous surfaces of Delhi
city International Journal of Applied Earth Observation and Geoinformation, 19, 348 - 358
• Mohr, P. J.; Newell, D. B. & Taylor, B. N., 2015. CODATA Recommended Values of the Fundamental
Physical Constants: 2014 National Institute of Standards and Technology, Committee on Data for Science
and Technology
• Moran, M.; Jackson, R.; Slater, P. & Teillet, P., 1992. Evaluation of simplified procedures for retrieval of
land surface reflectance factors from satellite sensor output Remote Sensing of Environment, 41, 169-184
• NASA, 2020. What is Synthetic Aperture Radar?. Available at https://earthdata.nasa.gov/learn/what-is-sar
Basic Tutorials
The following are very basic tutorials for land cover classification using the Semi-Automatic Classification Plugin
(SCP). It is assumed that you have a basic knowledge of QGIS (you can find a guide to QGIS interface at this
page).
5.1 Tutorial 1
The following is a basic tutorial about the land cover classification using the Semi-Automatic Classification Plugin
(SCP). It is assumed that you have a basic knowledge of QGIS.
This is a basic tutorial about the use of SCP for the classification of a multispectral image. It is recommended to
read the Brief Introduction to Remote Sensing (page 123) before this tutorial.
The purpose of the classification is to identify the following land cover classes:
1. Water;
2. Built-up;
155
Semi-Automatic Classification Plugin Documentation, Release 7.5.5.1
3. Vegetation;
4. Soil.
The study area of this tutorial is Matera (Basilicata, Italy) which is a city in the Southern Italy. Matera is the
site of a Center of the Italian Space Agency (ASI) specialized in remote sensing, which has several purposes, in
particular to acquire, process and store remote sensing data.
Following the video of this tutorial.
http://www.youtube.com/watch?v=Ceyhm3DlZNY
We are going to download a Sentinel-2 Satellite (page 128) image (Copernicus land monitoring services) and use
the bands illustrated in the following table.
TIP : In case of slow internet connection you can download an image subset from this archive (about
30 MB, © Copernicus Sentinel data 2020 downloaded from https://scihub.copernicus.eu/), unzip the
downloaded file, and skip to Preprocess the data (page 160).
Start QGIS and the SCP. Open the tab Download products (page 45) clicking the button in the Home
(page 21), or in the SCP menu (page 19),
In the tab Download products (page 45) click the button to display the OpenStreetMap tiles (© Open-
StreetMap contributors) in the QGIS map, licensed as CC BY-SA (Tile Usage Policy ).
We are searching a specific image acquired on 8 July 2020 because it is cloud free (however you can use any other
Sentinel-2 image). In Search parameters (page 48) enter the point coordinates:
• UL: 16.5 / 40.7
• LR: 16.7 / 40.6
TIP : In general it is possible to define the area coordinates clicking the button , then left
click in the map for the UL point and right click in the map for the LR point.
Select Sentinel-2 from the Products and set:
• Date from: 2020-07-08
• to: 2020-07-08
Now click the button Find and after a few seconds the image will be listed in the Product list (page 48).
Click the ProductID L2A_T33TXF_A026345_20200708T094035 in the table to display a preview that is
useful for assessing the quality of the image and the cloud cover.
We can also select which bands to be downloaded according to our purpose. In particular, select the tab Download
options (page 49) and check only the Sentinel-2 bands (that will be used in this tutorial): 2, 3, 4, 5, 6, 7, 8, 8A, 11,
12 and the ancillary data.
For the purpose of this tutorial, uncheck the option Preprocess images (you should usually leave this checked)
because we are going to preprocess the image in Preprocess the data (page 160). To start the image download,
click the button RUN and select a directory where bands are saved. The download could last a few minutes
according to your internet connection speed. The download progress is displayed in a bar.
TIP : The option Only if preview in Layers allows for downloading only images in the result
table which are loaded as previews in the map. If this option is unchecked, all the products in the list
are downloaded.
After the download, all the bands are automatically loaded in the map.
For for limiting the study area (and reducing the processing time) we can clip the image.
First, we need to define a Band set containing the bands to be clipped. Several tools in SCP use the Band sets,
which are numbered sets of raster bands and related information (such as center wavelength and acquisition date).
Open the tab Band set (page 31) clicking the button in the SCP menu (page 19) or the SCP dock (page 20).
Click the button to refresh the layer list, and select the bands: 2, 3, 4, 5, 6, 7, 8, 8A, 11, and 12; then click
In Preprocessing (page 50) open the tab Clip multiple rasters (page 62). We are going to clip the Band set 1 which
contains Sentinel-2 bands.
Click the button and select an area such as the following image (left click in the map for the UL point and
right click in the map for the LR point), or enter the following values:
• UL: 623830 / 4505410
• LR: 645330 / 4494350
Click the button RUN and select a directory where clipped bands are saved. New files will be created with
the file name prefix defined in Output name prefix. When the process is completed, clipped rasters are loaded and
displayed.
The downloaded product is a Sentinel-2 Level 2A image, which is already atmospherically corrected (Level 1C
images require atmospherical correction as described in Image conversion to reflectance (page 148)). Neverthe-
less, preprocessing is required to convert pixel values to a decimal value of reflectance.
Conversion to reflectance (see Radiance and Reflectance (page 126)) can be performed automatically. The meta-
data file (a .txt file whose name contains MTL) downloaded with the images contains the required information for
the conversion.
In order to convert bands to reflectance, open the Preprocessing (page 50) clicking the button in the SCP
menu (page 19) or the SCP dock (page 20), and select the tab Sentinel-2 (page 59).
Click the button Directory containing Sentinel-2 bands and select the directory of clipped Sentinel-2 bands.
The list of bands is automatically loaded in the table Metadata (page 60).
You can ignore Select metadata file because this conversion doesn’t require additional information.
Sentinel-2 Level 2A images don’t require the DOS1 Correction (page 149); if it was a Level 1C image we should
For the purpose of this tutorial, uncheck the option Create Band set and use Band set tools because we are
going to define this in the following step Define the Band set and create the Training Input File (page 163).
In order to start the conversion process, click the button RUN and select the directory where converted
bands are saved.
After a few minutes, converted bands are loaded and displayed (file name beginning with RT_). If Play
sound when finished is checked in Classification process (page 111) settings, a sound is played when the process
is finished.
We can remove all the bands loaded in QGIS layers except the ones whose name begin with RT_.
Define the Band set and create the Training Input File
Now we need to define the Band set which is the input image for SCP classification. Open the tab Band set
(page 31) clicking the button in the SCP menu (page 19) or the SCP dock (page 20).
In Band set definition click the button to clear all the bands from active band set created during the previous
steps.
Click the button to refresh the layer list, and select all the converted bands; then click to add selected
rasters to the Band set.
In the table Band set definition order the band names in ascending order (click to sort bands by name
automatically). Finally, select Sentinel-2 from the list Wavelength quick settings, in order to set automatically the
Center wavelength of each band and the Wavelength unit (required for spectral signature calculation).
We can display a Color Composite (page 133) of bands: Near-Infrared, Red, and Green: in the Working tool-
bar (page 27), click the list RGB= and select the item 7-3-2 (corresponding to the band numbers in Band set
(page 31)). You can see that image colors in the map change according to the selected bands, and vegetation is
highlighted in red (if the item 3-2-1 was selected, natural colors would be displayed).
TIP : If a Band set (page 31) is defined, a temporary virtual raster (named Virtual Band Set
1) is created automatically, which allows for the display of Color Composite (page 133). In order to
speed up the visualization, you can show only the virtual raster and hide all the layers in the QGIS
Layers.
Now we need to create the Training input (page 23) in order to collect Training Areas (page 133) (ROIs) and
calculate the Spectral Signature (page 126) thereof (which are used in classification).
In the SCP dock (page 20) select the tab Training input (page 23) and click the button to create the Training
input (define a name such as training.scp). The path of the file is displayed and a vector is added to QGIS
layers with the same name as the Training input (in order to prevent data loss, you should not edit this layer using
QGIS functions).
We are going to create ROIs defining the Classes and Macroclasses (page 134). Each ROI is identified by a Class
ID (i.e. C ID), and each ROI is assigned to a land cover class through a Macroclass ID (i.e. MC ID).
Macroclasses are composed of several materials having different spectral signatures; in order to achieve good
classification results we should separate spectral signatures of different materials, even if belonging to the same
macroclass. Thus, we are going to create several ROIs for each macroclass (setting the same MC ID, but assigning
a different C ID to every ROI).
We are going to used the Macroclass IDs defined in the following table.
Macroclasses
ROIs can be created by manually drawing a polygon or with an automatic region growing algorithm.
In the map zoom over the dark blue area in the lower left corner of the image which is a water body. To manually
create a ROI inside the dark area, click the button in the Working toolbar (page 27). Left click on the map to
define the ROI vertices and right click to define the last vertex closing the polygon. An orange semi-transparent
polygon is displayed over the image, which is a temporary polygon (i.e. it is not saved in the Training input).
TIP : You can draw temporary polygons (the previous one will be overridden) until the shape covers
the intended area.
If the shape of the temporary polygon is good we can save it to the Training input.
Open the Training input (page 23) to define the Classes and Macroclasses (page 134) . In the ROI & Signature
list (page 24) set MC ID = 1 and MC Name = Water; also set C ID = 1 and C Name = Lake. Now click to
save the ROI in the Training input.
After a few seconds, the ROI is listed in the ROI & Signature list (page 24) and the spectral signature is calculated
in the Working toolbar (page 27) and click over the light blue area of the map. After a while the orange
semi-transparent polygon is displayed over the image.
TIP : Dist value should be set according to the range of pixel values; in general, increasing this value
creates larger ROIs.
In the ROI & Signature list (page 24) set MC ID = 2 and MC Name = Built-up ; also set C ID = 2 (it should be
already set) and C Name = Buildings.
Again, the C ID in ROI & Signature list (page 24) is automatically increased by 1.
After clicking the button in the Working toolbar (page 27) you should notice that the cursor in the map
displays a value changing over the image. This is the NDVI value of the pixel beneath the cursor (NDVI is
displayed because the function Display is checked in ROI options (page 27)). The NDVI value can be useful
for identifying spectrally pure pixels, in fact vegetation has higher NDVI values than soil.
For instance, move the mouse over a vegetation area and left click to create a ROI when you see a local maximum
value. This way, the created ROI and the spectral signature thereof will be particularly representative of healthy
vegetation.
Create a ROI for the class Vegetation (red pixels in color composite RGB=7-3-2) and a ROI for the class
Soil (bare soil or low vegetation) (yellow pixels in color composite RGB=7-3-2) following the
same steps described previously. The following images show a few examples of these classes identified in the map.
Spectral signatures are used by Classification Algorithms (page 135) for labelling image pixels. Different materials
may have similar spectral signatures (especially considering multispectral images) such as built-up and soil. If
spectral signatures used for classification are too similar, pixels could be misclassified because the algorithm is
unable to discriminate correctly those signatures. Thus, it is useful to assess the Spectral Distance (page 141)
of signatures to find similar spectral signatures that must be removed. Of course the concept of distance vary
according to the algorithm used for classification.
Fig. 16: A temporary ROI created with the automatic region growing algorithm
Fig. 18: NDVI value of vegetation pixel displayed in the map. Color composite RGB = 7-3-2
One can simply assess spectral signature similarity by displaying a signature plot. In order to display the signature
plot, in the ROI & Signature list (page 24) highlight two or more spectral signatures (with click in the table), then
click the button . The Spectral Signature Plot (page 112) is displayed in a new window. Move and zoom
inside the Plot (page 114) to see if signatures are similar (i.e. very close). Double click the color in the Plot
Signature list (page 113) to change the line color in the plot.
We can see in the following figure a signature plot of different materials.
In the plot we can see the line of each signature (with the color defined in the ROI & Signature list (page 24)), and
the spectral range (minimum and maximum) of each band (i.e. the semi-transparent area colored like the signature
line). The larger is the semi-transparent area of a signature, the higher is the standard deviation, and therefore the
heterogeneity of pixels that composed that signature. Spectral similarity between spectral signatures is highlighted
in orange in the Plot Signature list (page 113).
Additionally, we can calculate the spectral distances of signatures (for more information see Spectral Distance
(page 141)). Highlight two or more spectral signatures with click in the table Plot Signature list (page 113), then
click the button ; distances will be calculated for each pair of signatures. Now open the tab Spectral distances
(page 116); we can notice that similarity between signatures vary according to considered algorithm.
For instance, two signatures can be very similar for Spectral Angle Mapping (page 137) (very low Spectral Angle
(page 142)), but quite distant for the Maximum Likelihood (page 136) (Jeffries-Matusita Distance (page 141) value
near 2). The similarity of signatures is affected by the similarity of materials (in relation to the number of spectral
bands available); also, the way we create ROIs influences the signatures.
Spectral signature values, standard deviation and other details such as the number of ROI pixels are displayed in
the Signature details (page 116).
We need to create several ROIs (i.e. spectral signatures) for each macroclass (repeating the steps in Create
the ROIs (page 165)), assigning a unique C ID to each spectral signature, and assess the spectral distance thereof
It is worth mentioning that you can show or hide the temporary ROI clicking the button ROI in Working
toolbar (page 27).
The classification process is based on collected ROIs (and spectral signatures thereof). It is useful to create a
Classification preview (page 30) in order to assess the results (influenced by spectral signatures) before the final
classification. In case the results are not good, we can collect more ROIs to better classify land cover.
Before running a classification (or a preview), set the color of land cover classes that will be displayed in the
classification raster. In the ROI & Signature list (page 24), double click the color (in the column Color) of each
ROI to choose a representative color of each class. Also, we need to set the color for macroclasses in ROI &
Signature list (page 24).
Now we need to select the classification algorithm. In this tutorial we are going to use the Maximum Likelihood
(page 136).
Open the tool Classification (page 70) to set the use of classes or macroclasses. Check Use C ID and in
Algorithm (page 71) select the Spectral Angle Mapping. The input band set is 1 because it is the number of the
band set containing the image (bands) that we want to classify.
buttons and for better displaying the Input image (i.e. image stretching).
In Classification preview (page 30) set Size = 300; click the button and then left click a point of the image in
the map. The classification process should be rapid, and the result is a classified square centered in clicked point.
Previews are temporary rasters (deleted after QGIS is closed) placed in a group named Class_temp_group in
the QGIS panel Layers. Now in Classification (page 70) check Use MC ID and click the button in
Classification preview (page 30).
We can see that now there are only 4 colors representing the macroclasses.
TIP : When loading a previously saved QGIS project, a message could ask to handle missing layers,
which are temporary layers that SCP creates during each session and are deleted afterwards; you can
click Cancel and ignore these layers; also, you can delete these temporary layers clicking the button
Assuming that the results of classification previews were good (i.e. pixels are assigned to the correct class defined
in the ROI & Signature list (page 24)), we can perform the actual land cover classification of the whole image.
In Classification (page 70) check Use MC ID. In the Classification output (page 72) click the button
and define the path of the classification output, which is a raster file (.tif). If Play sound when finished is
checked in Classification process (page 111) settings, a sound is played when the process is finished.
Well done! You have just performed your first land cover classification.
However, you can see that there are several classification errors, because the number of ROIs (spectral signatures)
is insufficient.
We can improve the classification using some of the tools that will be described in other tutorials.
5.2 Tutorial 2
The following is a basic tutorial about the Semi-Automatic Classification Plugin (SCP). It is assumed that you
have a basic knowledge of QGIS.
This is a basic tutorial about the use of SCP for the unsupervised classification of a multispectral image. It is
recommended to read the Brief Introduction to Remote Sensing (page 123) before this tutorial, and in particular
the part Clustering (page 146).
Clustering can be used for unsupervised classification, which means that no training input is required, producing
classes (i.e. clusters) that have no definition and consequently the user must assign a land cover label to each class.
The purpose of the classification is to identify the land cover classes with the corresponding ID codes defined in
the following table.
Classes
Input Data
Any multispectral image can be used for unsupervised classification. In this tutorial, we are going to use a subset
of a Sentinel-2 Satellite (page 128) image (Copernicus land monitoring services) and use the bands illustrated in
the following table. You can download the image from this archive (about 30 MB, © Copernicus Sentinel data
2020 downloaded from https://scihub.copernicus.eu/), and then unzip the downloaded file.
convert bands to reflectance, open the Preprocessing (page 50) clicking the button in the SCP menu (page 19)
or the SCP dock (page 20), and select the tab Sentinel-2 (page 59).
Click the button Directory containing Sentinel-2 bands and select the directory of Sentinel-2 bands. The list
of bands is automatically loaded in the table Metadata (page 60).
You can ignore Select metadata file because this conversion doesn’t require additional information.
Check the option Create Band set and use Band set tools in order to create automatically the input Band set
that will be used for clustering.
To start the conversion process, click the button RUN and select the directory where converted bands are
saved. After a few minutes, converted bands are loaded and displayed (file name beginning with RT_).
Open the tab Band set (page 31) clicking the button in the SCP menu (page 19) or the SCP dock (page 20).
The bands are already listed in the first Band set.
Clustering
Clustering requires the user to define the number of clusters (i.e. classes) expected in the image, which corre-
spond to as many spectral signatures (i.e. seeds). During the first iteration clusters are produced calculating the
pixel spectral distance with initial spectral signatures. Pixels are assigned according to the most similar spectral
signature, therefore producing clusters. In this tutorial we are going to perform clustering using the ISODATA
algorithm, which includes steps for merging clusters having similar spectral signatures and splitting clusters hav-
ing too high variability. We are going to select randomly the starting spectral signatures and use the algorithm
Euclidean Distance (page 142) for distance calculation.
Open the tab Clustering (page 73) clicking the button in the SCP menu (page 19) or the SCP dock (page 20).
In Select input band set we set 1 because we are going to classify the first Band set.
In Method check ISODATA. In Distance threshold set 0.01, which is the distance required for merging
clusters.
In Number of classes set 10; we set a number higher than the land cover classes defined at the beginning of this
tutorial because we try to account for spectral variability of classes. In Max number of iterations set 10; in general,
the more the iterations the better the result.
In ISODATA max standard deviation we set the maximum standard deviation considered for splitting a class; we
can set the value 0.2 here. In ISODATA minimum class size in pixels we can set 10.
We check the option Use random seed signatures to randomly selects the spectral signatures of seed pixels from
Band set. Now click the button RUN and define the path of the classification output.
After the process, the result is displayed in the map. The result of the unsupervised classification is a raster having
values that don’t correspond to the class IDs defined at the beginning of this tutorial. The number of classes can
vary depending on the various iterations.
The mean spectral signature of output classes is displayed in the panel. We could also save these spectral signatures
in a previously created Training input (page 23) if we wanted to use these in other supervised classifications.
We must identify the correspondence between raster classes and the land cover class IDs defined at the beginning
of this tutorial.
We can perform this task through photointerpretation. We can display a Color Composite (page 133) of bands:
Near-Infrared, Red, and Green: in the Working toolbar (page 27), click the list RGB= and select the item 7-3-2
(corresponding to the band numbers in Band set (page 31)). You can see that image colors in the map change
according to the selected bands, and vegetation is highlighted in red (if the item 3-2-1 was selected, natural
colors would be displayed).
We should take note of the corresponding values, similarly to the following table as example. Please note that you
should define different values according to the result of your classification output.
Now we can reclassify the output raster according to out classification system.
Open the tool Reclassification (page 88) clicking the button in the SCP menu (page 19) or the SCP dock
(page 20). Click Calculate unique values to automatically calculate unique values in the classification and
fill the reclassification table. In New value set the corresponding New class ID.
When the table is complete, click the button RUN and define the path of the classification output. After the
process, the raster is added to the map. We can define a symbology according to our classes.
Of course, the result of an unsupervised classification can have many classification errors. In this example, several
pixels of soil have been classified as built-up. For reducing those errors we could repeat the clustering step
increasing the Number of classes, reducing the value ISODATA max standard deviation, and increasing the value
Distance threshold.
The classification could contain small patches of pixels that we would like to remove from the classification result.
We can use the tool Classification sieve (page 82) to remove small patches of isolated pixel replacing them with
the value of the largest neighbour patch. Open the tool Classification sieve (page 82) clicking the button in
the SCP menu (page 19) or the SCP dock (page 20).
In Select the classification select the raster reclassified in the previous step (click to refresh the layer
list if the classification is not listed). In Size threshold enter 2 to set the maximum size of the patch to be replaced
(in pixel unit); all patches smaller than the selected number of pixels will be replaced by the value of the largest
neighbour patch. Setting the value 2 we are going to remove isolated single pixels.
In Pixel connection : select 8 as the type of pixel connection in a 3x3 window (i.e. diagonal pixels are
considered connected). Now click the button RUN and define the path of the classification output. After
the process, the raster is added to the map. We should see that single pixels have been replaced by the largest
neighbour patch, resulting in a cleaner classification.
Well done! We have performed an unsupervised classification of a remote sensing image. We can improve the
classification using some of the tools that will be described in other tutorials.
Thematic Tutorials
The following are thematic tutorials. Before these tutorials, it is recommended to read the Basic Tutorials
(page 155).
For other unofficial tutorials, also in languages other than English, see Other tutorials about SCP, also in languages
other than English (page 208).
191
Semi-Automatic Classification Plugin Documentation, Release 7.5.5.1
Semi-Automatic OS
To be updated.
The Semi-Automatic OS is a lightweight virtual machine for the land cover classification of remote sensing im-
ages. It includes the Semi-Automatic Classification Plugin (SCP) for QGIS, already configured along with all the
required dependencies.
The Semi-Automatic OS is based on Debian , and it is designed to require very little hardware resources. It
uses LXQT as main desktop environment. This virtual machine can be useful for testing the Semi-Automatic
Classification Plugin, or when the installation of the required programs in the host system is problematic.
The Semi-Automatic OS is available as 64 bit virtual machine that can be run in the open source VirtualBox, or
any other virtualization program. The following is a guide for the installation of the Semi-Automatic OS in the
open source program of virtualization VirtualBox.
193
Semi-Automatic Classification Plugin Documentation, Release 7.5.5.1
1. Download VirtualBox open source software (select a proper version depending on your OS) and install it;
at the end of the installation restart the system;
2. Download the Semi-Automatic OS virtual machine from here ;
3. Extract the virtual machine content in a directory; the file is compressed in 7z format (if needed, download
the open source extraction software from http://www.7-zip.org/);
4. Run VirtualBox and create a new Debian virtual machine;
1. Click the New button;
2. Type a name for the virtual machine (for instance Semi-Automatic OS); select Linux and Debian (32
or 64 bit) as Type and Version respectively; click Next;
3. Set the memory size; the more is the better, but this parameter should not exceed a half of the host
system RAM (for instance if the host system has 1 GB of RAM, type 512 MB); click Next;
4. In the Hard drive settings select Use an existing virtual hard drive file and select the downloaded file
SemiAutomaticOS.vmdk; click Create;
See http://www.gnu.org/licenses/.
197
Semi-Automatic Classification Plugin Documentation, Release 7.5.5.1
The SCP can be installed manually (this can be useful when an internet connection is not available, or the instal-
lation is required on multiple computers), following a few steps:
1. download the SCP zip archive from https://github.com/semiautomaticgit/
SemiAutomaticClassificationPlugin/archive/master.zip ;
2. extract the content of the archive (several files such as COPYING.txt and folders such as ui) in a new
folder named SemiAutomaticClassificationPlugin (without -master);
3. open the QGIS plugins directory (in Windows usually C:\Users\username\AppData\Roaming\QGIS\QGIS3\p
in Linux and Mac usually /home/username/.local/share/QGIS/QGIS3/profiles/
default/python/plugins) and delete the folder SemiAutomaticClassificationPlugin if
present;
4. copy the folder SemiAutomaticClassificationPlugin inside the QGIS plugins directory;
5. the plugin should be installed; start QGIS, open the Plugin Manager and be sure that Semi-Automatic
Classification Plugin is checked.
8.1.2 How to install the plugin from the official SCP repository?
It is possible to install the SCP using the official repository. This repository allows for the installation of the latest
version of SCP (master), in some cases also before the availability thereof in the QGIS repository. Therefore,
this can be useful if you need a fix or a new function that is still not available in the QGIS repository. Moreover, the
master version in the SCP repository can be installed along with the version available in the QGIS repository.
In order to install the SCP repository follow these steps:
• Run QGIS;
• From the main menu, select Plugins > Manage and Install Plugins;
SCP
URL:
https://semiautomaticgit.github.io/SemiAutomaticClassificationPlugin/repository.xml
• After the repository update, the item Semi-Automatic Classification Plugin - master
should be listed with the other plugins;
• From the menu All, select the Semi-Automatic Classification Plugin - master and
click the button Install plugin; the latest version of SCP should be automatically activated (ignore
errors, the restart of QGIS could be necessary to complete the SCP installation); it is possible to deactivate
the other SCP installed in the QGIS repository;
It is possible to install the previous version 6 of SCP using the official repository. Also, this version can be installed
along with the SCP version 7.
In order to install the SCP repository follow these steps:
• Run QGIS;
• From the main menu, select Plugins > Manage and Install Plugins;
URL:
https://semiautomaticgit.github.io/SemiAutomaticClassificationPlugin_v6/repository.
˓→xml
• After the repository update, the item Semi-Automatic Classification Plugin - master
should be listed with the other plugins;
• From the menu All, select the Semi-Automatic Classification Plugin - master and
click the button Install plugin; the latest version of SCP should be automatically activated (ignore
errors, the restart of QGIS could be necessary to complete the SCP installation); it is possible to deactivate
the other SCP installed in the QGIS repository;
In general, it is preferable to avoid thermal infrared bands. If you are using Landsat 4, 5 or 7 you should select
bands: 1, 2, 3, 4, 5, 7 avoiding band 6 that is thermal infrared; for Landsat 8 you should select bands: 2, 3, 4, 5,
6, 7. Landsat 8 band 1 is generally avoided because it is very similar to the blue band and it is mainly used for
coastal aerosol study. Landsat thermal infrared band is excluded from classifications because values are mainly
related to object temperature.
For Sentinel-2 images you can use bands: 2, 3, 4, 5, 6, 7, 8, 8A, 11, 12.
All Landsat 1,2, and 3 MSS and Landsat 4, 5, 7, and 8 images downloaded from http://earthexplorer.usgs.gov/ and
processed with the Level 1 Product Generation System (LPGS) can be converted to reflectance automatically by
the SCP; products generated by the LPGS have a MTL file included that is required for the conversion.
8.2.3 Can I apply the conversion to Sentinel-2 images download from the web?
Yes, you can convert also images downloaded from the web (actually the conversion is recommended). You
should move all the bands (.jp2 files) and if available the .xml file whose name contains MDT_SAFL1C in the
same directory. Then select this directory in Sentinel-2 conversion (page 60). Images are converted to reflectance.
8.2.4 Can I apply the conversion to Sentinel-2 L2A download from the web?
Yes, you should move all the .jp2 files inside the same directory and rename the files with the band number in
the ending of the name (e.g. from name_02_10m.jp2 to name_02.jp2) Then select this directory in Sentinel-2
conversion (page 60). Images are converted to reflectance.
8.2.5 How are converted Sentinel-2 images that have different resolutions?
During the conversion to reflectance, pixels of 20m bands are split in 4 pixels of 10m whose values are the same as
the original 20m pixel. The purpose of this operation is to allow for the calculation between all the bands, without
changing original values.
8.2.6 Can I apply the Landsat conversion and DOS correction to clipped bands?
Yes, you can clip the images before the conversion to reflectance and then copy the MTL file (contained in the
Landsat dataset) inside the directory with the clipped bands. If you want to apply the DOS correction (which
is an image based technique) you should convert the original Landsat bands (the entire image) and then clip the
conversion output (i.e. bands converted to reflectance).
8.2.7 Can I apply the DOS correction to bands with black border (i.e. with NoData
value)?
If you want to apply the DOS correction to an entire band which has NoData values (the black border with value =
0) then you have to check the checkbox Use value as NoData and set the value to 0. This is because DOS
is an image based technique, and NoData values must be excluded from the calculation.
DOS1 correction does not remove clouds from the image. However, Landsat 8 images include Band 9 that identi-
fies clouds (see this NASA site). You can use this band for the creation of a mask.
Also, see the following video-tutorial.
8.2.9 After pan-sharpening of Landsat 8 images, why NIR bands still have 30m
resolution?
Landsat 8 panchromatic band doesn’t acquire in the Near Infrared (NIR) region (see Landsat Satellites (page 127)).
Therefore, the pan-sharpening process can’t improve the resolution of NIR and SWIR bands (see Pan-sharpening
(page 145)), which appear to have 30m resolution. However, all pan-sharpened rasters have 15m resolution to
allow raster calculation.
8.3 Processing
Several materials have similar spectral signatures (e.g. soil and built-up, or forest and other types of dense low
vegetation), which can cause classification errors if ROIs, and spectral signatures thereof, are not acquired cor-
rectly. In order to improve the results, you can try to collect more ROIs over these areas, in order to train the
algorithm for these very similar areas, also, display the spectral signatures of these areas in Spectral Signature
Plot (page 112) to assess their similarity. You can also use a Signature threshold (page 43) for these signatures in
order to reduce the variability thereof (only pixels very similar to the input signatures will be classified). The Land
Cover Signature Classification (page 139) is also useful for classifying specific materials that can be spectrally
similar to other ones.
8.3.2 Is it possible to use the same training input for multiple images?
Yes, it is possible if all the images have the same number of bands. However, if images are acquired in different
months, land cover changes (especially of vegetation state) will affect the spectral signature (i.e. the same pixel has
different spectral signature in different periods). Atmospheric effects could also affect the images differently. That
could reduce classification accuracy. Therefore, it is suggested to collect always the ROIs and spectral signatures
for every image.
8.3.4 Can I use SCP with images from drones or aerial photographs?
Yes, you can use them if they have at least 4 bands. With less than 4 bands, semi-automatic classification al-
gorithms are unable to classify the land cover correctly. Alternative classification methods exist, such as object
oriented classification, which is not implemented in SCP.
8.3.5 Why using only Landsat 8 band 10 in the estimation of surface tempera-
ture?
Several methods were developed for estimating surface temperature. The method described in the tutorial for
temperature estimation requires only one band. Moreover, USGS recommends that users refrain from relying on
Landsat 8 Band 11 data in quantitative analysis of the Thermal Infrared Sensor data (see Changes to Thermal
Infrared Sensor (TIRS) data by USGS).
8.4 Warnings
8.4.1 Warning [12]: The following signature will be excluded if using Maximum
Likelihood. Why?
The ROI is too small (or too homogeneous) for the Maximum Likelihood (page 136) algorithm because that ROI
has a singular covariance matrix. You should create larger ROIs or don’t use the Maximum Likelihood algorithm
in the classification process.
8.5 Errors
If you found an error of the Semi-Automatic Classification Plugin please follow these steps in order to collect the
required information (log file):
1. close QGIS if already open;
2. open QGIS, open the Plugin tab Debug (page 108) and check the checkbox Records events in a log file
;
3. click the button Test dependencies in the tab Debug (page 108) ;
4. load the data in QGIS (or open a previously saved QGIS project) and repeat all the steps that cause the error in the Pl
• if the issue could be related to the image data, please use this sample dataset ;
5. if an error message appears (like the one in the following image), copy the whole content of the message in
a text file;
6. open the tab Debug (page 108) and uncheck the checkbox Records events in a log file, then click the
button and save the log file (which is a text file containing information about the Plugin processes);
7. open the log file and copy the whole content of the file;
Fig. 1: Debug
8. join the Facebook group , create a new post and copy the error message and the log file (or attach them).
The automatic creation of the virtual raster after Landsat conversion to reflectance is not required for the classi-
fication. Errors could happen if the output destination path contains special characters (such as accented letters)
or spaces; try to rename directories (e.g. rename new directory to new_directory). If you still get the
same error you can create a virtual raster manually.
It could be related to a wrong installation. Please, uninstall QGIS and install it again with administrative rights.
Then run QGIS and try to install the plugin following the Plugin Installation (page 3) guide.
Also, it could be related to the user name containing special characters. Please try the installation creating a new
user without special characters (e.g. user).
8.5.4 Error [50] ‘Internet error’. Unable to download Sentinel-2 images. Why?
The error message usually includes some information about the issue. First, check the user name and password.
If the account registration was recent, it could take a few days to complete the process for allowing the download
from SCP.
Also, there could be an interruption of the service. For Sentinel-2 images please check this website https://scihub.
copernicus.eu/news/ for messages about the state of the service.
In case you still get the same error, please follow these steps How can I report an error? (page 204).
The plugin requires the installation of GDAL, NumPy, SciPy and Matplotlib, which should be installed along with
QGIS. If the plugin installation fails, and you get a message about possible missing dependencies, you should
try to install or update QGIS and the required dependencies. Notice that in order to avoid this error, python
dependencies should not be installed through Anaconda.
8.6 Various
SCP allows for the land cover classification of remote sensing images through Supervised Classification
(page 133). You can produce a land cover raster using one of the Classification Algorithms (page 135) available in
SCP. These algorithms require spectral signatures or ROIs as input (for definitions please read Brief Introduction
to Remote Sensing (page 123)) that define the land cover classes to be identified in the image.
SCP can work with multispectral images acquired by satellites, airplanes, or drones. Also, SCP allows for the
direct search and download of free images (see Download products (page 45)). You cannot use orthophotos with
less than 4 bands, SAR data, and LIDAR data with SCP.
Input image in SCP is called Band set (page 31), which is used as input for the classification. SCP provides
several tools for the Preprocessing (page 50) of downloaded images, such as the conversion to reflectance and
manipulation of bands.
Classification results can be assessed with the tools Accuracy (page 77) and Classification report (page 81).
Also, rasters can be manipulated using Postprocessing (page 77) tools such as Classification to vector (page 81),
Reclassification (page 88), Edit raster (page 85) directly, Classification sieve (page 82), Classification erosion
(page 80), and Classification dilation (page 79).
The Spectral Signature Plot (page 112) and Scatter Plot (page 117) allow for the analysis of spectral signatures
and ROIs. Also, several Basic tools (page 34) are available for easing the ROI creation and editing spectral
signatures.
Raster calculation is available through the seamless integration of the tool Band calc (page 90) with bands in
the Band set (page 31), calculating mathematical expressions and spectral indices. Also, an output raster can be
calculated based on Decision rules (page 98).
The tool Batch (page 99) allows for the automatic execution of several SCP functions using a scripting interface.
See the Basic Tutorials (page 155) for more information and examples.
You can contribute to SCP by fixing and adding functionalities (see Where is the source code of SCP? (page 209)),
or translating the user manual (see How can I translate this user manual to another language? (page 208)).
Also, you can donate to this project at the following link https://fromgistors.blogspot.com/p/donations.html .
8.6.3 Free and valuable resources about remote sensing and GIS
8.6.4 Other tutorials about SCP, also in languages other than English
There are several tutorials about SCP on the internet. Following an incomplete list of these resources (please note
that these resources could use older versions of SCP):
• English: Webinar by NASA ARSET about Land Cover Classification with Satellite Imagery;
• English: Recommended Practice by UN-SPIDER about Burn Severity Mapping Burn Severity with QGIS;
• English: Using the SCP QGIS plugin to download and process Sentinel 2 imagery;
• French: Télédétection des feux de forêts;
• German: 2015 Jakob Erfassung von Landnutzungsveränderungen mit FOSS Image Processing Tools;
• Indonesian: Koreksi Radiometrik Menggunakan QGIS:Semi Automatic Classification;
• Italian: Classificazione e Mosaico di Varie Immagini Landsat;
• Korean: QGIS Semi-Automatic Classification Plugin;
• Portuguese: Workshop de Deteção Remota e Processamento Digital de Imagem com o QGIS e o Semi-
Automatic Classification Plugin;
• Portuguese: Classificacao supervisionada de imagens Sentinel-2 com QGIS e SCP;
• Portuguese: Avaliação do erro de uma imagem de satélite usando o QGIS e o SCP;
• Portuguese: Conversão Sentinel-2 para refletância com QGIS SCP;
• Portuguese: Criar composições coloridas no QGIS com SCP;
• Portuguese: Corte de imagem Sentinel-2 usand QGIS e SCP;
• Portuguese: Deteção Remota Com O QGIS;
• Russian: Sentinel- 2a Semi-Automatic Classification Plugin QGIS;
• Spanish: Descarga de imágenes de satélite desde servidores públicos;
• Swedish: Landsat 8 och fjärranalys med QGIS;
• Ukrainian: Semi-Automatic Classification 5.0;
• Ukrainian: Qgis ;
It is possible to easily translate the user manual to any language, because it is written in reStructuredText as
markup language (using Sphinx). Therefore, your contribution is fundamental for the translation of the manual to
your language. The following guide illustrates the main steps for the translation, which can be performed:
• using the free online service Transifex;
• using the gettext .po files.
Before translating, please read this document from the QGIS translation guide, which helps you understand the
reStructuredText.
Method 1. Translation using the free online service Transifex
This is probably the easiest way to translate the manual using an online service.
1. Join the Semi-automatic Classification Manual project
Go to the page https://www.transifex.com/semi-automatic-classification/
semi-automatic-classification-plugin-manual and click the button Help translate.
You can sign in using your Google or Facebook account, or with a free registration.
2. Select your language
Select your language and click the button Join team. If your language is not listed, click the
button Request language.
3. Translation
There are several files to be translated, which refer to the sections of the
SCP documentation. To translate the SCP interface you should select the file
semiautomaticclassificationplugin.ts .
Method 2. Translation using the gettext .po files
In order to use this method, you should be familiar with GitHub. This translation method allows for the translation
of the PO files locally.
1. Download the translation files
Go to the GitHub project https://github.com/semiautomaticgit/
SemiAutomaticClassificationManual_v4/tree/master/locale and download the .po files of
your language (you can add your language, if it is not listed), or you can fork the repository.
Every file .po is a text file that refers to a section of the User Manual.
2. Edit the translation files
Now you can edit the .po files. It is convenient to edit those file using one of the following
programs: for instance Poedit for Windows and Mac OS X, or Gtranslator for Linux or OmegaT
(Java based) for Windows, Linux and Mac OS X. These editors allow for an easy translation of
every sentence in the User Manual.
You can find the source code of SPC is at the following link https://github.com/semiautomaticgit/
SemiAutomaticClassificationPlugin